Statistics and Data Distributions

Data Types and Sampling Methods are two foundation ideas in statistics. Before we can analyze data, we need to know what kind of data we have and how it was collected. These choices affect what graphs make sense, what calculations are appropriate, and whether the conclusions are trustworthy.

In this lesson, you will learn how to distinguish categorical data from quantitative data, and how to evaluate four common sampling methods: simple random sampling, stratified sampling, cluster sampling, and systematic sampling. You will also learn how to spot possible bias in a sample.

Why this matters: If the wrong type of data is identified, the analysis can be misleading. If the sample is collected poorly, even perfect calculations will not give reliable results. Good statistics starts with good data.

1. Populations, Samples, and Bias

A population is the entire group you want to study. A sample is the smaller group actually observed or measured.

• Population: all 12th Grade students in a school
• Sample: 100 students chosen from that school

Because it is often too expensive or time-consuming to study every member of a population, statisticians use samples. The goal is for the sample to represent the population well.

Bias happens when a sampling method tends to favor certain outcomes or certain groups, causing the sample to not fairly represent the population.

For example, if a school surveys only students in advanced math classes about homework time, the results may not represent all students. That sample is likely biased because it excludes many students with different experiences.

2. Data Types

Statistics often begins by asking: What kind of variable is being recorded? At this level, the two main types are categorical and quantitative.

Categorical Data

Categorical data places individuals into groups or categories. The values are labels, not measurements. You do not add or average category names.

Examples of categorical data include:

• Eye color
• Favorite subject
• Type of phone used
• Whether a student has a part-time job: yes or no

For categorical data, we often summarize using:

• Counts
• Percentages
• Bar graphs
• Pie charts

Quantitative Data

Quantitative data consists of numerical values that represent amounts or measurements. These values usually make sense to add, subtract, average, or compare by size.

Examples of quantitative data include:

• Height in centimeters
• Test score
• Number of siblings
• Time spent studying

For quantitative data, we often summarize using:

• Mean or median
• Range or standard deviation
• Dot plots, histograms, or box plots

Important idea: Just because data uses numbers does not always mean it is quantitative. If the numbers are only labels, then the data is still categorical.

For example, if jerseys are numbered 4, 12, and 30, those numbers identify players, but they do not measure an amount. Jersey number is categorical, not quantitative.

3. Comparing Categorical and Quantitative Data

• Categorical: describes which group something belongs to
• Quantitative: describes how much or how many

You can ask yourself these questions:

1. Is this value a label or a measurement?
2. Would it make sense to calculate an average?
3. Does the variable place people into groups, or measure a number?

If the variable is a label or group, it is categorical. If it measures an amount, it is quantitative.

4. Worked Example 1: Identifying Data Types

Classify each variable as categorical or quantitative:

• a) Number of hours a student sleeps each night
• b) Blood type
• c) ZIP code
• d) Number of books read this month

Solution

• a) Number of hours a student sleeps each night: quantitative, because it measures an amount of time.
• b) Blood type: categorical, because it places people into groups such as A, B, AB, or O.
• c) ZIP code: categorical, because although it uses numbers, the numbers are labels for areas, not measurements.
• d) Number of books read this month: quantitative, because it counts how many books were read.

5. Sampling Methods

Once we know what data we want, we need to decide how to select the sample. Different sampling methods are designed for different situations. A good method reduces bias and helps the sample reflect the population.

A. Simple Random Sample

In a simple random sample, every possible sample of a given size has an equal chance of being selected. A common way to do this is to number all members of the population and use a random number generator.

Example: A principal wants to survey 50 students out of 1200 students. Each student is assigned a number from 1 to 1200, and 50 numbers are randomly chosen.

Why it is useful: It is one of the fairest sampling methods because it uses chance rather than human choice.

Possible limitation: It can be hard to carry out if a complete list of the population is not available.

B. Stratified Sample

In a stratified sample, the population is first divided into smaller groups called strata. These groups are formed based on an important characteristic, such as grade level, gender, or region. Then a simple random sample is taken from each stratum.

Example: A school wants a survey sample that represents all grade levels. Students are grouped into 9th, 10th, 11th, and 12th grades, then a random sample is taken from each grade.

Why it is useful: It ensures all important groups are represented.

Possible limitation: If the strata are chosen poorly, or if some strata are sampled unfairly, bias can still occur.

C. Cluster Sample

In a cluster sample, the population is divided into groups called clusters, usually based on location or natural grouping. Then one or more entire clusters are randomly selected, and everyone in those clusters is sampled.

Example: A district wants to survey students. It randomly chooses 3 schools and surveys every student in those schools.

Why it is useful: It is often cheaper and easier, especially when the population is spread out.

Possible limitation: If the chosen clusters are not representative of the whole population, the sample may be biased.

D. Systematic Sample

In a systematic sample, members of the population are selected according to a fixed rule, usually every \(k\)-th person after a random starting point.

If a population has size \(N\) and you want a sample of size \(n\), a common interval is

$$k \approx \frac{N}{n}$$

Then you choose a random starting number from 1 to \(k\), and select every \(k\)-th individual.

Example: A manager has a list of 500 customers and wants a sample of 50. Since

$$k = \frac{500}{50} = 10,$$

the manager randomly chooses a starting point from 1 to 10, then selects every 10th customer on the list.

Why it is useful: It is simple and fast.

Possible limitation: If there is a pattern in the list that matches the sampling interval, the sample may be biased.

6. How to Tell the Sampling Methods Apart

• Simple random: choose individuals directly at random from the whole population
• Stratified: split into important groups, then sample from each group
• Cluster: split into natural groups, randomly choose whole groups, then survey everyone in them
• Systematic: choose every \(k\)-th person after a random start

A good shortcut is this:

• If every group is sampled, it is usually stratified.
• If only some groups are chosen and everyone in them is surveyed, it is usually cluster.

7. Worked Example 2: Identifying Sampling Methods

Identify the sampling method in each situation.

a) A researcher assigns every student in a college a number and uses a random number generator to select 200 students.

b) A company divides workers into departments and randomly selects 10 workers from each department.

c) A city divides neighborhoods into blocks, randomly selects 5 blocks, and surveys every household on those blocks.

d) A teacher chooses a random starting student on the class list, then surveys every 3rd student.

Solution

• a) Simple random sample
• b) Stratified sample, because the population is divided into departments and each department is sampled
• c) Cluster sample, because whole blocks are selected and everyone in them is surveyed
• d) Systematic sample, because every 3rd student is chosen after a random start

8. Bias in Sampling

Even if a method sounds scientific, it can still produce a biased sample. We should always ask: Does this method give all parts of the population a fair chance to be represented?

Common sources of bias include:

• Undercoverage: some groups are left out of the sampling process
• Convenience sampling: choosing people who are easy to reach instead of choosing randomly
• Voluntary response: only people who choose to respond are included
• Poor timing or wording: when data collection favors certain answers

For this concept, focus especially on whether the sampling method may leave out part of the population or overrepresent one group.

Examples of possible bias:

• Surveying only athletes about school lunch may not represent all students.
• Taking every 10th name from a list arranged by team membership may accidentally over-sample one group.
• Choosing one neighborhood cluster to represent an entire city may be misleading if that neighborhood is unusual.

9. Worked Example 3: Assessing Bias

A school wants to know how much time students spend on homework each night. A survey is given only to students in the library after school.

Question: Is this likely to be biased?

Solution

Yes, this is likely biased. Students in the library after school may be more likely to study longer than the average student. The sample does not fairly represent all students in the school.

This is not one of the four standard random methods. It is closer to a convenience sample because the school surveyed students who were easy to reach.

10. Worked Example 4: Choosing and Evaluating a Method

A district wants to estimate the average number of hours students spend on screens per day. The district has elementary, middle, and high school students. Which sampling method would be a good choice: simple random, stratified, cluster, or systematic?

Solution

A stratified sample would be a strong choice. The district can divide students into elementary, middle, and high school groups, then randomly sample from each group.

This is a good idea because screen time may differ by age group. If the district used only a simple random sample, one school level might be underrepresented by chance. Stratified sampling helps ensure that all major groups are included.

A cluster sample could be easier, such as choosing a few schools and surveying all students there, but it may be less accurate if those schools are not typical of the whole district.

11. Quick Comparison Table in Words

• Categorical data: group or label
• Quantitative data: number or measurement
• Simple random sample: random individuals from the whole population
• Stratified sample: random individuals from every group
• Cluster sample: random whole groups, then survey everyone in them
• Systematic sample: every \(k\)-th individual after a random start
• Bias: a sample that does not fairly represent the population

12. Tips for Test Questions

• If the values are words or labels, think categorical.
• If the values measure an amount, think quantitative.
• If the problem says “randomly choose people from each group,” think stratified.
• If it says “randomly choose groups and survey everyone in them,” think cluster.
• If it says “every 5th,” “every 10th,” or similar, think systematic.
• If it says “use a random number generator on the whole population,” think simple random.
• Always ask whether some people are more likely than others to be included. If so, there may be bias.

Summary

In statistics, categorical data places individuals into groups, while quantitative data measures numerical amounts. Correctly identifying the data type helps determine how the data should be summarized and displayed.

To collect data fairly, statisticians use sampling methods such as simple random, stratified, cluster, and systematic sampling. Each method has strengths, but each can also lead to bias if used poorly. A good statistical conclusion depends on both the right kind of data and a representative sample.

Experimental Design vs. Observational Studies is a key idea in statistics because it helps us decide what conclusions we are allowed to make from data.

Many students see two variables connected in a dataset and immediately think one variable caused the other. But in statistics, a relationship in data does not always mean cause and effect. To understand the difference, we need to know how the data were collected.

This lesson will explain the difference between observational studies and experiments, and show how ideas like randomization, control groups, placebos, and lurking variables help us judge whether a claim is about correlation or causation.

1. Correlation vs. Causation

Correlation means that two variables are associated. When one changes, the other tends to change too.

For example, students who spend more time studying may tend to earn higher test scores. That is a correlation between study time and test score.

Causation means that changes in one variable directly produce changes in another variable.

If increasing study time actually leads to higher scores, then study time is causing the increase in scores.

The big statistical question is: Can we really say one variable causes the other, or can we only say they are related?

In general:

• Observational studies can show correlation.
• Well-designed experiments can provide evidence for causation.

2. What Is an Observational Study?

In an observational study, researchers observe individuals and measure variables, but they do not assign treatments or try to change who gets what.

They simply record what is already happening.

Examples of observational studies include:

• Surveying teenagers about how many hours they sleep and their grades.
• Recording whether adults exercise regularly and whether they have high blood pressure.
• Looking at past records to compare smoking habits and lung disease.

Because the researcher does not assign treatments, there may be other differences between the groups besides the variable being studied.

That means an observational study can show that two variables move together, but it usually cannot prove causation.

3. What Is an Experiment?

In an experiment, the researcher actively imposes a treatment on some individuals and then compares results.

A treatment is the condition applied to individuals, such as taking a medicine, using a new teaching method, or following a special exercise plan.

Experiments are stronger than observational studies because the researcher controls how treatments are assigned.

If the experiment is designed well, especially with random assignment, then differences in outcomes are more likely to be caused by the treatment itself rather than by other factors.

4. Key Features of a Good Experiment

To decide whether an experiment can support a causal conclusion, look for these important features.

A. Random Assignment

Random assignment means placing subjects into treatment groups by chance.

For example, if 100 patients are part of a medicine study, each patient might be randomly assigned to either:

• a treatment group that receives the medicine, or
• a control group that does not receive the medicine.

Random assignment helps create groups that are similar overall. This reduces the effect of other variables and makes the comparison fairer.

B. Control Group

A control group is the group that does not receive the main treatment, or receives a standard treatment for comparison.

The control group gives us a baseline. Without it, we would not know whether the treatment actually made a difference.

C. Placebo

A placebo is a fake treatment that looks like the real treatment but has no active effect.

For example, in a medicine study, one group might receive the actual pill while the control group receives a pill with no medicine in it.

This matters because people sometimes improve simply because they believe they are being treated. This is called the placebo effect.

D. Blinding

Blinding means that people do not know which treatment was given.

• Single-blind: the subjects do not know whether they got the real treatment or placebo.
• Double-blind: neither the subjects nor the researchers directly interacting with them know who got which treatment.

Blinding helps prevent expectations from affecting the results.

E. Replication

Replication means using enough subjects in the experiment. A larger number of subjects helps make results more reliable and less likely to happen just by chance.

5. Lurking Variables and Confounding

A lurking variable is a variable that is not included in the study but affects the variables being studied.

These variables can make it seem like one variable causes another when that may not be true.

For example, suppose a study finds that students who carry water bottles get better grades. Does carrying a water bottle cause better grades? Probably not.

A lurking variable might be organization. More organized students may be both more likely to bring a water bottle and more likely to earn high grades.

When groups differ in an important way besides the treatment, we say there is confounding. Confounding makes it hard to tell which variable is really causing the effect.

6. How to Tell Whether a Study Supports Causation

When you read a study description, ask these questions:

1. Did researchers assign a treatment, or did they just observe existing behavior?
2. Was there random assignment to groups?
3. Was there a control group?
4. Was a placebo used if appropriate?
5. Could there be lurking variables or confounding?

If researchers only observed people, then the study usually shows association only.

If researchers used a randomized experiment with strong controls, then a cause-and-effect conclusion is much more reasonable.

7. Important Difference: Random Sampling vs. Random Assignment

Students often confuse these two ideas.

• Random sampling helps us choose a sample that represents a larger population.
• Random assignment helps us create similar treatment groups in an experiment.

Random sampling supports generalizing results to a population.

Random assignment supports causal conclusions.

A study may have one, both, or neither.

8. Worked Example 1: Basic Classification

A researcher surveys 500 adults and records how many cups of coffee each person drinks per day and how many hours they sleep each night.

Question: Is this an observational study or an experiment?

Solution:

The researcher is only collecting information and not assigning people to drink a certain amount of coffee. That means this is an observational study.

Conclusion: The researcher may find a correlation between coffee drinking and sleep, but cannot prove that coffee caused the sleep difference from this study alone.

9. Worked Example 2: Recognizing an Experiment

A school wants to know whether a new online math program improves test scores. They randomly assign 60 students to two groups:

• 30 students use the new online program.
• 30 students use the usual textbook method.

After 8 weeks, the school compares the test scores.

Question: Is this an observational study or an experiment? Can the school discuss causation?

Solution:

This is an experiment because the school assigned students to treatments.

It also used random assignment, which helps make the groups similar at the beginning.

The textbook group is the control group because it provides the comparison condition.

Conclusion: If the online-program group scores higher, the school has good evidence that the program caused the improvement, assuming the experiment was carried out fairly.

10. Worked Example 3: Lurking Variable

A study finds that teenagers who play more sports tend to have higher grades.

Question: Does this prove that playing sports causes higher grades?

Solution:

No. This may be an observational study, so it shows a relationship but not necessarily causation.

A possible lurking variable is family support. Students with strong family support may be more likely to join sports and also more likely to do well in school.

Another lurking variable might be time management. Students who manage time well may both participate in sports and keep up their grades.

Conclusion: We can say there is an association, but we cannot conclude that sports participation alone caused higher grades.

11. Worked Example 4: Evaluating Experimental Design

A company tests a new energy drink. Volunteers are split into two groups. Group A receives the new drink. Group B receives a similar-looking drink without the active ingredient. The volunteers do not know which drink they received. After one week, reaction times are measured.

Question: What features of good experimental design are present?

Solution:

• Group A receives the treatment.
• Group B is the control group.
• The similar-looking inactive drink is a placebo.
• Because volunteers do not know which drink they received, the study uses blinding.

If the volunteers were also assigned to groups by chance, then the design would include random assignment, making a causal conclusion stronger.

Conclusion: This study has several strong experimental design features that help reduce bias and support cause-and-effect reasoning.

12. Quick Comparison Table

• Observational Study: observes subjects, does not assign treatment, can show correlation, cannot usually prove causation.
• Experiment: imposes treatment, often uses random assignment, can provide evidence for causation.

13. Common Mistakes to Avoid

• Do not say caused when the data came from an observational study.
• Do not assume a strong correlation means one variable must cause the other.
• Do not forget about lurking variables.
• Do not confuse random sampling with random assignment.
• Do not ignore the importance of a control group or placebo in an experiment.

14. Final Summary

To tell whether a study supports correlation or causation, first look at how the data were collected.

An observational study only records what happens naturally, so it can show an association between variables but usually cannot prove one causes the other.

An experiment applies treatments to subjects. If it uses random assignment, a control group, and sometimes a placebo and blinding, it gives much stronger evidence for causation.

Always watch for lurking variables, because they can create misleading relationships in data. In statistics, the difference between “related” and “caused by” is extremely important.

Measures of Center and Spread help us describe a data set in two important ways:

• Center: Where the data tends to cluster
• Spread: How much the data varies

These ideas are very important in statistics because a single number is usually not enough to describe a data set well. Two data sets can have the same average but very different amounts of variation.

In this lesson, you will learn how to compute and interpret the mean, median, mode, range, interquartile range (IQR), variance, and standard deviation.

1. Measures of Center

Measures of center describe a typical value in a data set.

A. Mean

The mean is the average. To find it, add all the data values and divide by the number of values.

If the data values are \(x_1, x_2, x_3, \dots, x_n\), then the mean is

$$\bar{x} = \frac{x_1+x_2+x_3+\cdots+x_n}{n}$$

The mean uses every value in the data set, which makes it useful. However, it can be strongly affected by very large or very small values called outliers.

B. Median

The median is the middle value when the data is arranged in order.

• If there is an odd number of data values, the median is the middle one.
• If there is an even number of data values, the median is the average of the two middle values.

The median is often a better measure of center when the data has outliers or is skewed, because it is less affected by extreme values.

C. Mode

The mode is the value that appears most often.

A data set can have:

• One mode (unimodal)
• Two modes (bimodal)
• More than two modes
• No mode if no value repeats

The mode is especially useful for categorical data or when you want to know the most common value.

2. Measures of Spread

Measures of spread describe how far apart the data values are.

A. Range

The range is the difference between the greatest value and the least value.

$$\text{Range} = \text{maximum} - \text{minimum}$$

The range is simple to calculate, but it depends only on two values, so it can be affected a lot by outliers.

B. Interquartile Range (IQR)

The interquartile range describes the spread of the middle 50% of the data.

To find the IQR:

1. Order the data from least to greatest.
2. Find the median.
3. Find \(Q_1\), the median of the lower half.
4. Find \(Q_3\), the median of the upper half.
5. Subtract:

$$\text{IQR} = Q_3 - Q_1$$

The IQR is resistant to outliers, so it is useful when the data is skewed.

C. Variance

The variance measures how far the data values tend to be from the mean.

To calculate it, we first find each value's distance from the mean, square those distances, and then average them.

For a population, variance is written as

$$\sigma^2 = \frac{\sum (x-\mu)^2}{N}$$

For a sample, variance is written as

$$s^2 = \frac{\sum (x-\bar{x})^2}{n-1}$$

In many school problems, you will be told whether the data is a population or a sample. If not, many classroom examples use the sample formula.

The reason for squaring is that positive and negative deviations from the mean would otherwise cancel out.

D. Standard Deviation

The standard deviation is the square root of the variance.

For a population:

$$\sigma = \sqrt{\frac{\sum (x-\mu)^2}{N}}$$

For a sample:

$$s = \sqrt{\frac{\sum (x-\bar{x})^2}{n-1}}$$

Standard deviation is often easier to interpret than variance because it is in the same units as the original data.

A small standard deviation means the data values are close to the mean. A large standard deviation means the data values are more spread out.

3. Choosing the Best Measure

Different situations call for different measures.

• Use mean and standard deviation when the data is fairly symmetric and has no strong outliers.
• Use median and IQR when the data is skewed or has outliers.
• Use mode when you want the most common value.
• Use range for a quick idea of total spread, but remember it is sensitive to extreme values.

4. Worked Examples

Example 1: Finding mean, median, mode, and range

The quiz scores are: 6, 8, 7, 9, 8, 10, 8

Step 1: Order the data

\(6, 7, 8, 8, 8, 9, 10\)

Mean

$$\bar{x} = \frac{6+7+8+8+8+9+10}{7} = \frac{56}{7} = 8$$

Median

There are 7 values, so the middle value is the 4th value: \(8\)

Mode

The value \(8\) appears most often, so the mode is \(8\).

Range

$$10-6=4$$

Answer:

• Mean = \(8\)
• Median = \(8\)
• Mode = \(8\)
• Range = \(4\)

This data set is centered at 8, and the scores are spread over 4 points.

Example 2: Finding quartiles and IQR

The data set is: 3, 5, 7, 8, 12, 13, 15, 18, 21

The data is already in order.

Step 1: Find the median

There are 9 values, so the median is the 5th value: \(12\)

Step 2: Find the lower half and upper half

Lower half: \(3, 5, 7, 8\)

Upper half: \(13, 15, 18, 21\)

Step 3: Find \(Q_1\) and \(Q_3\)

\(Q_1\) is the median of \(3, 5, 7, 8\):

$$Q_1 = \frac{5+7}{2}=6$$

\(Q_3\) is the median of \(13, 15, 18, 21\):

$$Q_3 = \frac{15+18}{2}=16.5$$

Step 4: Find the IQR

$$\text{IQR} = Q_3-Q_1 = 16.5-6 = 10.5$$

Answer:

• Median = \(12\)
• \(Q_1 = 6\)
• \(Q_3 = 16.5\)
• IQR = \(10.5\)

This tells us that the middle 50% of the data lies within an interval of length 10.5.

Example 3: Finding sample variance and sample standard deviation

The data set is: 2, 4, 6, 8

Step 1: Find the mean

$$\bar{x} = \frac{2+4+6+8}{4} = \frac{20}{4}=5$$

Step 2: Find each deviation from the mean

\(2-5=-3\), \(4-5=-1\), \(6-5=1\), \(8-5=3\)

Step 3: Square each deviation

$$(-3)^2=9, \quad (-1)^2=1, \quad 1^2=1, \quad 3^2=9$$

Step 4: Add the squared deviations

$$9+1+1+9=20$$

Step 5: Divide by \(n-1\) for sample variance

Since \(n=4\), we divide by \(3\):

$$s^2 = \frac{20}{3} \approx 6.67$$

Step 6: Take the square root for standard deviation

$$s = \sqrt{\frac{20}{3}} \approx 2.58$$

Answer:

• Sample variance \(\approx 6.67\)
• Sample standard deviation \(\approx 2.58\)

This means the data values typically vary from the mean by about 2.58 units.

Example 4: Interpreting outliers and choosing good measures

Consider the two data sets:

Set A: \(72, 75, 78, 80, 82\)

Set B: \(72, 75, 78, 80, 120\)

Mean of Set A

$$\bar{x} = \frac{72+75+78+80+82}{5} = \frac{387}{5}=77.4$$

Median of Set A

The median is \(78\).

Mean of Set B

$$\bar{x} = \frac{72+75+78+80+120}{5} = \frac{425}{5}=85$$

Median of Set B

The median is still \(78\).

Notice what happened: one extreme value, \(120\), pulled the mean up from \(77.4\) to \(85\), but the median stayed the same.

Conclusion: For Set B, the median is a better measure of center because the data has an outlier.

5. Common Mistakes to Avoid

• Not ordering data before finding the median or quartiles
• Mixing up mean and median
• Forgetting to divide by \(n-1\) when using the sample variance formula
• Calling the largest repeated value the mode instead of the most frequent value
• Using range or mean alone to describe a data set completely

6. Key Ideas to Remember

• The mean is the average.
• The median is the middle value.
• The mode is the most common value.
• The range is maximum minus minimum.
• The IQR is \(Q_3-Q_1\), the spread of the middle 50%.
• The variance measures average squared distance from the mean.
• The standard deviation is the square root of variance and shows typical spread from the mean.

Brief Summary

Measures of center tell us what a typical value looks like, while measures of spread tell us how much the data varies. The most useful pair is often mean and standard deviation for symmetric data, or median and IQR for skewed data or data with outliers. When describing any data set, think about both the center and the spread.

Graphical Displays of Univariate Data are graphs used to show and interpret the distribution of one variable at a time. In 12th Grade statistics, the most common displays are histograms, box plots, and density curves.

These graphs help us answer important questions about a dataset:

• Where is the center of the data?
• How spread out are the values?
• Is the distribution symmetric or skewed?
• Are there one or more peaks?
• Are there possible outliers?

Learning to read these displays is important because raw data by itself can be hard to understand. A good graph helps us quickly see patterns that would be difficult to notice from a list of numbers.

Univariate data means data involving only one measured characteristic. For example, heights of students, test scores, hours of sleep, or daily temperatures are all univariate datasets.

When describing a distribution from a graph, a helpful checklist is:

• Shape: symmetric, skewed left, skewed right, uniform, or bimodal/multimodal
• Center: where the data cluster, often around the median or mean
• Spread: how variable the data are, such as range or interquartile range
• Outliers: unusually high or low values

This is often remembered as shape, center, spread, and outliers.

1. Histograms

A histogram groups numerical data into intervals called bins and shows how many values fall into each bin. The bars touch because the data are numerical and continuous or measured on a scale.

To construct a histogram:

1. Choose equal-width intervals.
2. Count how many data values fall in each interval.
3. Draw a bar for each interval with height equal to the frequency or relative frequency.

When interpreting a histogram, look for:

• Peaks: high bars show where values are concentrated
• Gaps: intervals with few or no values
• Skewness: whether one tail is longer than the other
• Modality: the number of clear peaks

A distribution is symmetric if the left and right sides look roughly like mirror images.

A distribution is skewed right if the tail extends farther to the right. This usually means a few unusually large values pull the distribution in that direction.

A distribution is skewed left if the tail extends farther to the left. This means a few unusually small values stretch the distribution in that direction.

A distribution is bimodal if it has two clear peaks. More generally, a distribution with more than one peak is called multimodal.

Example 1: Interpreting a histogram

Suppose a teacher records the number of hours 20 students studied for a test:

2, 3, 3, 4, 4, 4, 5, 5, 5, 5, 6, 6, 6, 7, 7, 8, 9, 10, 10, 12

Use bins of width 2:

• 2 to <4: 3 values
• 4 to <6: 7 values
• 6 to <8: 5 values
• 8 to <10: 2 values
• 10 to <12: 2 values
• 12 to <14: 1 value

If we drew the histogram, most of the data would be in the middle, around 4 to 7 hours, with a tail stretching toward larger values like 10 and 12.

Interpretation:

• Shape: skewed right
• Center: about 5 or 6 hours
• Spread: from 2 to 12 hours
• Possible outliers: 12 might be unusually large, but the histogram alone does not prove it

This example shows how a histogram gives an overall picture of the distribution.

Important note about histograms: the appearance can change depending on bin width. If bins are too wide, important features may be hidden. If bins are too narrow, the graph may look too rough or confusing. So a histogram should be interpreted carefully.

2. Box Plots

A box plot, also called a box-and-whisker plot, summarizes a dataset using the five-number summary:

• Minimum
• First quartile, \(Q_1\)
• Median
• Third quartile, \(Q_3\)
• Maximum

The box extends from \(Q_1\) to \(Q_3\). A line inside the box marks the median. The whiskers extend toward the smallest and largest non-outlier values.

The width of the box is called the interquartile range, or IQR:

$$IQR = Q_3 - Q_1$$

The IQR measures the spread of the middle 50% of the data.

Box plots are especially useful for:

• Finding possible outliers
• Comparing distributions quickly
• Showing spread and skewness

Finding outliers with the 1.5 IQR rule

A value is considered a possible outlier if it is:

$$\text{less than } Q_1 - 1.5(IQR)$$

or

$$\text{greater than } Q_3 + 1.5(IQR)$$

These are called the lower fence and upper fence.

Example 2: Constructing and interpreting a box plot

Consider the dataset:

4, 5, 7, 8, 9, 10, 12, 13, 15, 18, 25

The data are already in order. There are 11 values.

Step 1: Find the median

The middle value is the 6th value, so the median is \(10\).

Step 2: Find \(Q_1\) and \(Q_3\)

The lower half is:

4, 5, 7, 8, 9

Its median is \(7\), so \(Q_1 = 7\).

The upper half is:

12, 13, 15, 18, 25

Its median is \(15\), so \(Q_3 = 15\).

Step 3: Compute the IQR

$$IQR = 15 - 7 = 8$$

Step 4: Find the fences

$$Q_1 - 1.5(IQR) = 7 - 1.5(8) = 7 - 12 = -5$$

$$Q_3 + 1.5(IQR) = 15 + 1.5(8) = 15 + 12 = 27$$

Any value below \(-5\) or above \(27\) would be a possible outlier. Since all values lie between 4 and 25, there are no possible outliers.

Five-number summary:

• Minimum: 4
• \(Q_1 = 7\)
• Median: 10
• \(Q_3 = 15\)
• Maximum: 25

Interpretation:

• The middle 50% of the data are between 7 and 15.
• The distribution has a longer upper side than lower side, suggesting some right skew.
• There are no formal outliers by the 1.5 IQR rule.

How box plots show skewness

In a box plot, skewness is suggested when one whisker is much longer than the other, or when the median is not centered inside the box.

• If the right whisker is longer, the distribution may be skewed right.
• If the left whisker is longer, the distribution may be skewed left.
• If the box and whiskers are balanced, the distribution may be roughly symmetric.

However, box plots do not show every detail. They do not clearly show gaps or the exact number of peaks. For that reason, histograms and box plots are often used together.

Example 3: Identifying an outlier using a box plot rule

Consider the ordered data:

3, 4, 5, 5, 6, 6, 7, 8, 9, 10, 22

Step 1: Median

There are 11 values, so the median is the 6th value: \(6\).

Step 2: Quartiles

Lower half: 3, 4, 5, 5, 6

So \(Q_1 = 5\).

Upper half: 7, 8, 9, 10, 22

So \(Q_3 = 9\).

Step 3: IQR

$$IQR = 9 - 5 = 4$$

Step 4: Fences

$$Q_1 - 1.5(IQR) = 5 - 1.5(4) = 5 - 6 = -1$$

$$Q_3 + 1.5(IQR) = 9 + 1.5(4) = 9 + 6 = 15$$

Any value greater than 15 is a possible outlier. Since \(22 > 15\), the value 22 is a possible outlier.

Interpretation:

• Most of the data lie between 3 and 10.
• The value 22 is much larger than the rest.
• A box plot would show this clearly as an outlier beyond the upper whisker.
• The distribution is skewed right because of the high value.

3. Density Curves

A density curve is a smooth curve that represents the overall pattern of a distribution. It is not usually built from exact counts the way a histogram is. Instead, it gives a smoothed model of how the data are distributed.

A density curve helps us think about the shape of a distribution in a more general way.

There are two important facts about density curves:

• The curve is always on or above the horizontal axis.
• The total area under the curve is exactly \(1\).

This means area under the curve represents proportion or probability.

For example, if 40% of the area lies between \(a\) and \(b\), then about 40% of the data values are between \(a\) and \(b\).

Density curves are useful when discussing distributions such as the normal distribution, which is symmetric and bell-shaped.

Describing shape with density curves

• Symmetric density curve: left and right sides match
• Right-skewed density curve: long tail to the right
• Left-skewed density curve: long tail to the left
• Bimodal density curve: two peaks

Density curves do not show individual data values, but they are useful for seeing the general shape of a distribution.

Example 4: Interpreting a density curve

Suppose a density curve for quiz scores is bell-shaped and centered at 75.

This tells us:

• Most scores are near 75.
• Scores much lower or much higher than 75 are less common.
• The distribution is roughly symmetric.
• The mean and median are likely close to 75.

Now suppose another density curve has one peak near 60 and another near 90.

This suggests a bimodal distribution. One possible explanation is that the data come from two different groups, such as students who studied very differently or two different classes combined into one dataset.

Comparing the three displays

• Histogram: shows the overall shape, peaks, gaps, and skewness clearly
• Box plot: shows center, spread, quartiles, and possible outliers efficiently
• Density curve: shows a smooth model of the distribution’s shape

No single graph tells the whole story. For example, a box plot may suggest skewness but cannot clearly show whether a distribution is bimodal. A histogram can show bimodality, but it may not identify outliers as precisely as a box plot using the 1.5 IQR rule.

Common descriptive words for distributions

• Symmetric: balanced on both sides
• Skewed right: longer tail on the right
• Skewed left: longer tail on the left
• Unimodal: one peak
• Bimodal: two peaks
• Multimodal: more than one peak
• Uniform: values spread fairly evenly
• Outlier: value far from the rest of the data

How to write a good interpretation

When asked to interpret a graph, do more than just name the graph. Describe the distribution using complete statistical language.

For example, instead of writing:

“The graph goes up and then down.”

Write something like:

“The histogram is unimodal and skewed right, with a center around 18 and a possible high outlier near 35.”

This kind of answer is clearer and more mathematically accurate.

Common mistakes to avoid

• Confusing skewed right with “most data are on the right.” Skew direction is named for the tail, not where most data are.
• Assuming every unusual value is automatically an outlier. Use the graph and, for box plots, the 1.5 IQR rule.
• Forgetting that box plots do not show peaks or gaps well.
• Using a histogram with uneven bins unless the graph is designed very carefully.
• Describing center without considering shape. For skewed data, the median is often a better description of center than the mean.

Quick review of key ideas

• Use a histogram to see shape, skewness, gaps, and modality.
• Use a box plot to summarize data with quartiles and identify possible outliers.
• Use a density curve to represent the overall shape of a distribution smoothly.
• Describe distributions by shape, center, spread, and outliers.
• Use the rule $$IQR = Q_3 - Q_1$$ and the fences $$Q_1 - 1.5(IQR), \quad Q_3 + 1.5(IQR)$$ to check for possible outliers.

Summary

Graphical displays of univariate data help us organize and understand a single variable. Histograms show the overall shape of the data, box plots summarize the data using quartiles and highlight possible outliers, and density curves provide a smooth picture of the distribution.

When interpreting any of these graphs, always focus on shape, center, spread, and outliers. Also pay attention to skewness and whether the distribution is unimodal, bimodal, or multimodal. These ideas are the foundation for understanding data distributions in statistics.

Lesson: Z-Scores and Standardized Data

When we collect data, the numbers are often measured on different scales. For example, one test might be scored out of 100, while another might be based on running time in seconds. This can make direct comparisons difficult.

Z-scores solve this problem by turning raw data values into standardized values. A z-score tells us how many standard deviations a data value is above or below the mean.

This makes it easier to compare values from different data sets, understand how unusual a value is, and work with the normal distribution.

1. What a z-score means

The mean is the center of a data set, and the standard deviation tells us how spread out the data are. A z-score combines both ideas into one number.

The formula for a z-score is:

$$z = \frac{x-\mu}{\sigma}$$

where:

• x is the data value,
• \(\mu\) is the mean,
• \(\sigma\) is the standard deviation.

If you are working with a sample instead of a whole population, you may also see:

$$z = \frac{x-\bar{x}}{s}$$

where \(\bar{x}\) is the sample mean and \(s\) is the sample standard deviation.

In either case, the idea is the same: subtract the mean, then divide by the standard deviation.

2. Interpreting z-scores

A z-score tells both direction and distance from the mean.

• If \(z = 0\), the value is exactly at the mean.
• If \(z > 0\), the value is above the mean.
• If \(z < 0\), the value is below the mean.
• The larger the absolute value of the z-score, the farther the value is from the mean.

For example:

• \(z = 1\) means the value is 1 standard deviation above the mean.
• \(z = -2\) means the value is 2 standard deviations below the mean.
• \(z = 2.5\) means the value is 2.5 standard deviations above the mean.

3. Why standardizing data is useful

Standardizing data means converting raw values into z-scores. This is useful because it allows us to:

• compare scores from different distributions,
• identify unusually high or low values,
• connect data to the normal distribution,
• describe position in a distribution clearly.

Suppose one student scores 82 on a math test and 18 on a reaction-time task. Without more information, we cannot tell which performance is stronger. But if we convert both to z-scores, we can compare them fairly.

4. Worked Example 1: Finding a z-score

A student scores 85 on an exam. The exam scores have a mean of 70 and a standard deviation of 10. Find the student's z-score.

Step 1: Write the formula

$$z = \frac{x-\mu}{\sigma}$$

Step 2: Substitute the values

$$z = \frac{85-70}{10}$$

Step 3: Simplify

$$z = \frac{15}{10} = 1.5$$

Answer: The z-score is 1.5.

Interpretation: The student's score is 1.5 standard deviations above the mean.

5. Worked Example 2: A negative z-score

A basketball player scores 12 points in a game. The team's point totals have a mean of 18 and a standard deviation of 3. Find and interpret the z-score.

Step 1: Use the formula

$$z = \frac{x-\mu}{\sigma}$$

Step 2: Substitute

$$z = \frac{12-18}{3}$$

Step 3: Simplify

$$z = \frac{-6}{3} = -2$$

Answer: The z-score is -2.

Interpretation: The player scored 2 standard deviations below the mean.

This is lower than average, and because it is fairly far from the mean, it may be considered an unusually low game compared to the team's typical scoring.

6. Comparing values from different data sets

One of the most important uses of z-scores is comparing values that are measured differently.

To decide which performance is better, compare the z-scores, not the raw scores.

Worked Example 3: Comparing two test scores

A student takes two tests:

• Math test: score 88, mean 80, standard deviation 4
• Science test: score 92, mean 85, standard deviation 10

On which test did the student perform better relative to the class?

Math test z-score:

$$z = \frac{88-80}{4} = \frac{8}{4} = 2$$

Science test z-score:

$$z = \frac{92-85}{10} = \frac{7}{10} = 0.7$$

Compare the z-scores:

• Math: \(z = 2\)
• Science: \(z = 0.7\)

Answer: The student performed better on the math test relative to the class.

Even though 92 is a higher raw score than 88, the math score is farther above its mean when measured in standard deviations.

7. Finding a raw score from a z-score

Sometimes you know the z-score and need to find the original data value. Start with the formula:

$$z = \frac{x-\mu}{\sigma}$$

Solve for \(x\):

$$x = \mu + z\sigma$$

This formula helps you move from standardized data back to the original units.

Worked Example 4: Finding the original value

A test score has a z-score of 1.2. The test has a mean of 75 and a standard deviation of 5. Find the original score.

Step 1: Use the formula

$$x = \mu + z\sigma$$

Step 2: Substitute

$$x = 75 + (1.2)(5)$$

Step 3: Simplify

$$x = 75 + 6 = 81$$

Answer: The original score is 81.

Interpretation: A score of 81 is 1.2 standard deviations above the mean.

8. Z-scores and the normal distribution

Z-scores are especially helpful when data are approximately normally distributed. In a normal distribution, many values are near the mean, and fewer values are far from the mean.

When data follow a normal shape, z-scores help describe where values fall in the distribution.

• About 68% of values lie within 1 standard deviation of the mean.
• About 95% of values lie within 2 standard deviations of the mean.
• About 99.7% of values lie within 3 standard deviations of the mean.

This is called the 68-95-99.7 rule.

So if a value has a z-score of 3 or -3, it is very far from the mean and is quite unusual in a normal distribution.

9. What standardized data looks like

When a whole data set is converted to z-scores, the new standardized data set has special properties:

• the mean becomes 0,
• the standard deviation becomes 1.

This is useful because every standardized data set is put onto the same scale. That makes comparisons easier and more meaningful.

10. Common mistakes to avoid

• Using the wrong mean or standard deviation: Make sure you use the values from the correct data set.
• Forgetting the sign: If the score is below the mean, the z-score should be negative.
• Comparing raw scores instead of z-scores: Different scales require standardization before comparison.
• Mixing up unusual and impossible: A large z-score means a value is uncommon, not impossible.

11. Quick strategy for solving z-score problems

1. Identify the value \(x\), mean, and standard deviation.
2. Use the formula $$z = \frac{x-\mu}{\sigma}$$ if you need the z-score.
3. Use the formula $$x = \mu + z\sigma$$ if you need the original value.
4. Interpret the result in words: above or below the mean, and by how many standard deviations.
5. If comparing two performances, compare the z-scores directly.

12. Final summary

A z-score tells how far a value is from the mean in units of standard deviation. Positive z-scores are above the mean, negative z-scores are below the mean, and a z-score of 0 is exactly at the mean.

Standardized data allows fair comparisons between different data sets, even when the original units are different. Z-scores are also important for understanding normal distributions and identifying values that are unusually high or low.

If you remember the key formula $$z = \frac{x-\mu}{\sigma}$$ and can explain what the result means, you have the main idea of z-scores and standardized data.

Introduction

In statistics, many real-life measurements follow a pattern called the normal distribution. Heights, test scores, measurement errors, and many biological traits often form this shape when graphed.

The normal distribution is often called a bell-shaped curve because it is highest in the middle and tapers off evenly on both sides. It is also symmetric, which means the left and right sides are mirror images.

Understanding the normal distribution helps us answer questions like:

• What percent of values are close to the average?
• How unusual is a certain score?
• What score marks the top 10% or bottom 5%?

In this lesson, you will learn how to use the Empirical Rule for quick estimates and how to use standard normal tables or technology to find more exact probabilities and percentiles.

1. What is a normal distribution?

A normal distribution is a continuous probability distribution with these important features:

• It is bell-shaped.
• It is symmetric about the mean.
• The mean, median, and mode are equal and at the center.
• Most values are near the center, and fewer values are far away.

A normal distribution is described by two numbers:

• The mean \, \(\mu\), which tells the center.
• The standard deviation \, \(\sigma\), which tells the spread.

If the standard deviation is small, the curve is narrow and tall. If the standard deviation is large, the curve is wider and flatter.

2. Standard deviation and distance from the mean

The standard deviation measures how far values typically are from the mean. On a normal curve, we often mark points that are 1, 2, and 3 standard deviations away from the mean:

$$\mu - 3\sigma,\; \mu - 2\sigma,\; \mu - \sigma,\; \mu,\; \mu + \sigma,\; \mu + 2\sigma,\; \mu + 3\sigma$$

These marks help us estimate how much of the data lies in different parts of the curve.

3. The Empirical Rule: 68-95-99.7 rule

For distributions that are approximately normal, the Empirical Rule gives quick estimates:

• About 68% of the data lies within 1 standard deviation of the mean.
• About 95% of the data lies within 2 standard deviations of the mean.
• About 99.7% of the data lies within 3 standard deviations of the mean.

In symbols:

$$\begin{aligned} P(\mu-\sigma \le X \le \mu+\sigma) &\approx 0.68 \\ P(\mu-2\sigma \le X \le \mu+2\sigma) &\approx 0.95 \\ P(\mu-3\sigma \le X \le \mu+3\sigma) &\approx 0.997 \end{aligned}$$

Because the curve is symmetric, these percentages split evenly on both sides of the mean.

That means:

• From the mean to \(\mu+\sigma\) is about 34%.
• From \(\mu+\sigma\) to \(\mu+2\sigma\) is about 13.5%.
• From \(\mu+2\sigma\) to \(\mu+3\sigma\) is about 2.35%.
• Beyond \(\mu+3\sigma\) is about 0.15% on each side.

These come from splitting the total percentages:

• \(68\%\) in the middle leaves \(32\%\) outside, so \(16\%\) is in each tail beyond 1 standard deviation.
• \(95\%\) in the middle leaves \(5\%\) outside, so \(2.5\%\) is in each tail beyond 2 standard deviations.
• \(99.7\%\) in the middle leaves \(0.3\%\) outside, so \(0.15\%\) is in each tail beyond 3 standard deviations.

4. Standardizing with z-scores

To compare values from a normal distribution, we often convert them into z-scores. A z-score tells how many standard deviations a value is above or below the mean.

The formula is:

$$z=\frac{x-\mu}{\sigma}$$

Here:

• \(x\) is the value,
• \(\mu\) is the mean,
• \(\sigma\) is the standard deviation.

Interpretation:

• If \(z=0\), the value is exactly at the mean.
• If \(z=1\), the value is 1 standard deviation above the mean.
• If \(z=-2\), the value is 2 standard deviations below the mean.

Once a value is written as a z-score, we can use a standard normal table or technology to find exact probabilities.

5. The standard normal distribution

The standard normal distribution is a special normal distribution with:

$$\mu=0 \quad \text{and} \quad \sigma=1$$

Its variable is usually written as \(Z\). Every normal distribution can be changed into the standard normal distribution by using z-scores.

Standard normal tables usually give the area to the left of a z-score, written as \(P(Z \le z)\).

For example, if a table says

$$P(Z \le 1.25)=0.8944$$

that means about 89.44% of the data lies below \(z=1.25\).

6. Finding percentiles

A percentile tells the value below which a certain percent of the data falls.

• The 50th percentile is the median, which is also the mean for a normal distribution.
• The 90th percentile is the value with 90% of the data below it.
• The 25th percentile is the value with 25% of the data below it.

To find a percentile for a normal distribution:

1. Find the z-score that matches the desired left-tail area using a z-table or technology.
2. Convert back to the original units using

$$x=\mu+z\sigma$$

7. When should you use the Empirical Rule, and when should you use a table or technology?

Use the Empirical Rule when:

• the distribution is approximately bell-shaped, and
• the value is exactly 1, 2, or 3 standard deviations from the mean, or can be handled by combining those regions.

Use a z-table or technology when:

• you want a more exact answer,
• the z-score is not a whole number like 1 or 2,
• you need a percentile such as the 92nd percentile.

Worked Example 1: Using the Empirical Rule

A set of exam scores is approximately normal with mean \(\mu=70\) and standard deviation \(\sigma=8\).

Question: About what percent of students scored between 62 and 78?

Step 1: Identify how far each value is from the mean.

$$62=70-8=\mu-\sigma$$

$$78=70+8=\mu+\sigma$$

So the interval is within 1 standard deviation of the mean.

Step 2: Apply the Empirical Rule.

About 68% of the data lies between \(\mu-\sigma\) and \(\mu+\sigma\).

Answer: About 68% of students scored between 62 and 78.

Worked Example 2: Combining Empirical Rule regions

The heights of a group of plants are approximately normal with mean \(\mu=50\) cm and standard deviation \(\sigma=4\) cm.

Question: About what percent of plants are taller than 58 cm?

Step 1: Express 58 in terms of standard deviations.

$$58=50+2(4)=\mu+2\sigma$$

So we want the area to the right of \(\mu+2\sigma\).

Step 2: Use the Empirical Rule.

About 95% of the data lies within 2 standard deviations of the mean, so 5% lies outside.

Because the curve is symmetric, half of that 5% is in the upper tail.

$$\frac{5\%}{2}=2.5\%$$

Answer: About 2.5% of plants are taller than 58 cm.

Worked Example 3: Using a z-score and standard normal table

Suppose SAT-style scores are normally distributed with mean \(\mu=500\) and standard deviation \(\sigma=100\).

Question: What percent of students score below 650?

Step 1: Find the z-score.

$$z=\frac{x-\mu}{\sigma}=\frac{650-500}{100}=1.5$$

Step 2: Use the standard normal table or technology.

We need \(P(Z \le 1.5)\).

From a z-table,

$$P(Z \le 1.50)\approx 0.9332$$

Step 3: Interpret the result.

About 93.32% of students score below 650.

Answer: About 93.3%.

Note: The Empirical Rule could only give an estimate here because \(z=1.5\) is not exactly 1, 2, or 3.

Worked Example 4: Finding a percentile value

A manufacturer finds that the lifetimes of its light bulbs are normally distributed with mean \(\mu=1200\) hours and standard deviation \(\sigma=100\) hours.

Question: What lifetime marks the 90th percentile?

Step 1: Identify the z-score for the 90th percentile.

We want the z-score with 0.9000 area to the left.

From a z-table or technology,

$$z \approx 1.28$$

Step 2: Convert back to the original units.

$$x=\mu+z\sigma$$

$$x=1200+1.28(100)=1200+128=1328$$

Answer: The 90th percentile is about 1328 hours.

This means about 90% of the bulbs last less than 1328 hours, and about 10% last longer.

8. Common mistakes to avoid

• Using the Empirical Rule on a distribution that is not bell-shaped. The rule only works well for distributions that are approximately normal.
• Forgetting symmetry. The left and right sides of a normal distribution have equal area.
• Confusing percent below with percent above. If a table gives the area to the left, then the area to the right is

$$1-P(Z\le z)$$

• Mixing up z-score and x-value. A z-score is measured in standard deviations, while an x-value is in the original units.
• Not checking whether the answer makes sense. A value above the mean should usually have a positive z-score, and a value below the mean should usually have a negative z-score.

9. Quick strategy guide

When solving a normal distribution problem, ask yourself:

1. Is the distribution approximately normal?
2. Am I finding a percentage/probability or a value/percentile?
3. Should I use the Empirical Rule for an estimate, or a z-table/technology for an exact answer?
4. If needed, did I convert between \(x\) and \(z\) correctly?

Brief Summary

The normal distribution is a symmetric, bell-shaped distribution centered at its mean. The Empirical Rule tells us that about 68%, 95%, and 99.7% of the data lie within 1, 2, and 3 standard deviations of the mean.

For more exact probabilities and percentiles, we use z-scores and the standard normal distribution. A z-score shows how many standard deviations a value is from the mean, and z-tables or technology help us find exact areas under the curve.

By combining these tools, you can estimate percentages quickly, calculate more precise probabilities, and find percentile values in real-world contexts.

Bivariate Data and Scatterplots

In statistics, we often want to study how two quantitative variables are related. This type of data is called bivariate data. For example, we might compare hours studied and test score, height and shoe size, or temperature and ice cream sales.

A scatterplot is one of the most useful graphs for bivariate data. It helps us see whether two variables are connected, whether the relationship is positive or negative, and whether the pattern looks linear or not.

Understanding scatterplots is important because they are the starting point for describing relationships and later building regression models. Before we calculate an equation or make predictions, we first need to look at the data.

1. What is bivariate data?

Bivariate means “two variables.” Each individual or object in the data set has two numerical values recorded for it.

• A student: hours studied and exam score
• A city: average temperature and electricity use
• A car: age and resale value

Each pair of values can be written as an ordered pair, like \((x,y)\).

Usually:

• The explanatory variable goes on the horizontal axis, called the x-axis.
• The response variable goes on the vertical axis, called the y-axis.

For example, if we think study time affects test score, then:

• \(x =\) hours studied
• \(y =\) test score

2. What is a scatterplot?

A scatterplot is a graph where each data pair \((x,y)\) is plotted as a point on a coordinate plane.

The graph helps us answer questions like:

• Do the variables seem related?
• Is the association positive, negative, or neither?
• Does the pattern look roughly linear?
• Are there clusters or unusual points?

Unlike bar graphs or histograms, a scatterplot is designed specifically to show the relationship between two numerical variables.

3. How to read a scatterplot

When looking at a scatterplot, focus on four main features:

1. Direction
2. Form
3. Strength
4. Unusual features

Direction

Direction tells us whether the variables move together or in opposite ways.

• Positive association: as \(x\) increases, \(y\) tends to increase.
• Negative association: as \(x\) increases, \(y\) tends to decrease.
• No clear association: the points do not show an obvious upward or downward trend.

Examples:

• Hours studied and test score often show a positive association.
• Age of a car and resale value often show a negative association.

Form

Form describes the overall shape of the pattern.

• Linear: the points follow a pattern close to a straight line.
• Nonlinear: the points follow a curve or another shape.

In this lesson, we pay special attention to linear trends, because they are very common in Grade 12 statistics and lead into regression.

Strength

Strength tells us how closely the points follow the pattern.

• Strong association: points lie close to a line or curve.
• Weak association: points are more spread out.

If the points are tightly grouped around a line, the relationship is stronger. If they are widely scattered, the relationship is weaker.

Unusual features

Scatterplots can also show special patterns that matter:

• Clusters: groups of points that are close together
• Gaps: spaces where few or no points appear
• Outliers: points far from the general pattern

These features can affect how we interpret the relationship.

4. Positive, negative, and no association

Here is how to recognize the basic types of association.

Positive association

If the points go upward from left to right, the association is positive.

Example: Students who spend more time practicing math problems often earn higher quiz scores.

Negative association

If the points go downward from left to right, the association is negative.

Example: As the number of absences increases, course grade may decrease.

No clear association

If the points look scattered without a clear upward or downward pattern, there may be little or no association.

Example: A student’s shoe size and biology test score likely have no meaningful relationship.

5. Linear trends

A linear trend means the points lie roughly around a straight line. The line does not need to pass through every point. We are looking for an overall pattern.

If the relationship is linear, then as \(x\) changes by equal amounts, \(y\) tends to change by fairly consistent amounts.

For example, if a scatterplot of study time and test score shows points roughly along an upward-sloping line, that suggests a positive linear trend.

If a scatterplot of car age and value shows points roughly along a downward-sloping line, that suggests a negative linear trend.

6. Clustering in scatterplots

Sometimes points form clusters instead of one single pattern. A cluster is a group of points gathered in one area of the graph.

Clusters may happen because:

• the data come from different groups,
• there are different categories mixed together, or
• the relationship changes in different situations.

For example, a scatterplot of height and weight for teenagers and adults together may show two clusters because the two age groups differ.

When you see clusters, be careful about describing the graph as one simple trend. There may be more than one group hidden in the data.

7. Outliers and why they matter

An outlier is a point that lies far away from the rest of the data. In a scatterplot, an outlier may not fit the overall pattern.

Outliers matter because they can:

• suggest a recording error,
• show an unusual individual, or
• change the appearance of the relationship.

For example, if most students studied between 1 and 5 hours, but one student studied 10 hours and still scored very low, that point could be an outlier.

When describing a scatterplot, mention outliers if you see them.

8. Correlation is not causation

A scatterplot can show that two variables are associated, but that does not automatically mean one variable causes the other.

For example, ice cream sales and sunglasses sales may both increase together. That does not mean buying ice cream causes people to buy sunglasses. A third factor, such as warm weather, may affect both.

This is an important idea in statistics: association does not prove causation.

9. Steps for describing a scatterplot

When asked to interpret a scatterplot, use a clear process:

1. State the two variables being compared.
2. Describe the direction: positive, negative, or none.
3. Describe the form: linear or nonlinear.
4. Describe the strength: strong, moderate, or weak.
5. Mention any clusters, gaps, or outliers.

A strong answer might say:

“The scatterplot compares hours studied and test score. It shows a positive, roughly linear, moderate-to-strong association. As study time increases, test score tends to increase. There are no major outliers.”

10. Worked Examples

Example 1: Identifying direction

A teacher records the following data for students:

• Hours studied: \(1, 2, 3, 4, 5\)
• Test scores: \(58, 64, 71, 78, 84\)

The ordered pairs are:

$$ (1,58), (2,64), (3,71), (4,78), (5,84) $$

If we plot these points, they rise from left to right.

Interpretation:

• Direction: positive
• Form: roughly linear
• Strength: strong, because the points are close to a line

This suggests that students who studied more tended to score higher.

Example 2: Negative association

A used car dealer tracks car age and resale value:

• Age in years: \(1, 2, 3, 4, 5\)
• Value in thousands of dollars: \(22, 19, 17, 14, 12\)

The ordered pairs are:

$$ (1,22), (2,19), (3,17), (4,14), (5,12) $$

As the car gets older, its value decreases.

Interpretation:

• Direction: negative
• Form: roughly linear
• Strength: strong

This is a clear example of a negative linear trend.

Example 3: Recognizing an outlier

Suppose the data for hours studied and test score are:

$$ (1,55), (2,60), (3,68), (4,74), (5,80), (6,35) $$

The first five points show an upward trend, but the point \((6,35)\) is far below the pattern.

Interpretation:

• Overall direction: positive
• Form: roughly linear for most points
• Unusual feature: outlier at \((6,35)\)

This outlier could mean the student had a bad test day, the score was recorded incorrectly, or another factor affected the result.

Example 4: Recognizing clusters

A data set combines exercise time and heart rate for two groups:

$$ (10,90), (12,92), (14,95), (16,97), (30,120), (32,123), (35,127), (36,129) $$

If plotted, the first four points would form one group and the last four points would form another group.

Interpretation:

• There appears to be a positive association.
• The graph also shows two clusters.
• This may mean the data come from two different populations or situations.

Instead of treating all the points as one simple pattern, we should notice the clustering and think about why it appears.

11. Common mistakes to avoid

• Confusing slope with strength: a steep trend is not always stronger than a gentle trend. Strength depends on how closely the points follow the pattern.
• Ignoring outliers: unusual points can change the story of the graph.
• Assuming causation: two variables moving together does not prove one causes the other.
• Forgetting the variables: always name what is on the \(x\)-axis and \(y\)-axis.
• Describing only direction: a complete description includes direction, form, strength, and unusual features.

12. How this connects to regression

Scatterplots are often the first step before drawing a line of best fit or using linear regression. If the scatterplot shows a roughly linear pattern, then a regression line may be useful for modeling the relationship.

If the plot has no clear trend, strong curvature, or separate clusters, then a simple linear model may not be the best choice.

This is why visualizing the data is so important: the graph helps us decide what kind of model makes sense.

Brief Summary

Bivariate data involve two numerical variables recorded together for each individual. A scatterplot displays these pairs as points on a graph so we can study their relationship.

When interpreting a scatterplot, describe the direction of the association, the form of the pattern, the strength of the relationship, and any clusters, gaps, or outliers. Look especially for whether the data show a positive or negative linear trend.

Scatterplots help us understand data before making predictions or building regression models. They are a key tool for analyzing relationships between two variables in statistics.

Pearson Correlation Coefficient is a number that measures how strongly two variables are related in a linear way.

It is usually written as 7 or, more commonly in statistics, as r. In this lesson, we will use r.

If two variables tend to increase together, the correlation is positive. If one variable tends to decrease when the other increases, the correlation is negative. If there is no clear linear pattern, the correlation is close to 0.

The Pearson correlation coefficient helps us answer questions like these:

• Do students who study more tend to score higher on a test?
• As outdoor temperature rises, does ice cream sales increase?
• Is there a relationship between height and shoe size?

However, an important warning must always be remembered: correlation does not mean causation. Just because two variables are related does not prove that one causes the other.

1. What values can Pearson's r take?

The value of Pearson's correlation coefficient always lies between 1 and 1:

$$-1 \le r \le 1$$

• r = 1: perfect positive linear relationship
• r = -1: perfect negative linear relationship
• r = 0: no linear relationship

Most real data sets give a value somewhere between these extremes.

2. Interpreting the sign of r

The sign tells the direction of the relationship.

• If r > 0, the relationship is positive: as one variable increases, the other tends to increase.
• If r < 0, the relationship is negative: as one variable increases, the other tends to decrease.

For example, if study hours and test scores have a positive correlation, students who study more generally tend to score higher.

If speed and travel time for a fixed distance have a negative correlation, then greater speed usually means less time.

3. Interpreting the size of r

The absolute value of r, written as |r|, tells the strength of the linear relationship.

• Values of |r| close to 1 mean a strong linear relationship.
• Values of |r| close to 0 mean a weak linear relationship.

A common rough guide is:

• 0 to 0.3: weak
• 0.3 to 0.7: moderate
• 0.7 to 1.0: strong

This guide uses absolute value, so it applies to both positive and negative correlations. For example, r = -0.82 is a strong negative correlation.

4. What does Pearson's r actually measure?

Pearson's r measures how closely the points in a scatter plot follow a straight-line pattern.

It does not measure whether the relationship is curved. A data set can have a strong curved pattern and still have a correlation near 0.

So before interpreting r, it is always a good idea to look at a scatter plot.

5. Formula for Pearson's correlation coefficient

For paired data values (x_1,y_1), (x_2,y_2), \dots, (x_n,y_n), the Pearson correlation coefficient is:

$$r = \frac{\sum (x-\bar{x})(y-\bar{y})}{\sqrt{\sum (x-\bar{x})^2 \sum (y-\bar{y})^2}}$$

Here:

• \bar{x} is the mean of the x-values
• \bar{y} is the mean of the y-values
• (x-\bar{x}) is the deviation of an x-value from the mean
• (y-\bar{y}) is the deviation of a y-value from the mean

This formula compares how the x-values and y-values vary together.

If values above the mean for x often match values above the mean for y, the numerator tends to be positive, giving a positive correlation.

If values above the mean for x often match values below the mean for y, the numerator tends to be negative, giving a negative correlation.

6. Steps for calculating r by hand

When calculating Pearson's r by hand, use this process:

1. List the paired data.
2. Find the mean of x and the mean of y.
3. Compute (x-\bar{x}) and (y-\bar{y}) for each pair.
4. Find (x-\bar{x})(y-\bar{y}), (x-\bar{x})^2, and (y-\bar{y})^2.
5. Add each column.
6. Substitute into the formula.

In many classes, a calculator or technology is used for the final value. Even so, you should still understand what the number means.

Worked Example 1: A perfect positive correlation

Suppose the data are:

• x: 1, 2, 3, 4
• y: 2, 4, 6, 8

These points lie exactly on the line y = 2x, so we expect a perfect positive linear relationship.

Find the means:

$$\bar{x} = \frac{1+2+3+4}{4} = 2.5$$ $$\bar{y} = \frac{2+4+6+8}{4} = 5$$

Now make a table:

$$\begin{array}{c|c|c|c|c|c} x & y & x-\bar{x} & y-\bar{y} & (x-\bar{x})(y-\bar{y}) & (x-\bar{x})^2,(y-\bar{y})^2 \\\hline 1 & 2 & -1.5 & -3 & 4.5 & 2.25, 9 \\ 2 & 4 & -0.5 & -1 & 0.5 & 0.25, 1 \\ 3 & 6 & 0.5 & 1 & 0.5 & 0.25, 1 \\ 4 & 8 & 1.5 & 3 & 4.5 & 2.25, 9 \end{array}$$

Add the needed values:

$$\sum (x-\bar{x})(y-\bar{y}) = 10$$ $$\sum (x-\bar{x})^2 = 5$$ $$\sum (y-\bar{y})^2 = 20$$

Substitute into the formula:

$$r = \frac{10}{\sqrt{5 \cdot 20}} = \frac{10}{\sqrt{100}} = \frac{10}{10} = 1$$

Interpretation: There is a perfect positive linear correlation.

Worked Example 2: A perfect negative correlation

Suppose the data are:

• x: 1, 2, 3, 4
• y: 8, 6, 4, 2

As x increases, y decreases in a perfectly straight-line way.

Find the means:

$$\bar{x} = 2.5, \qquad \bar{y} = 5$$

Table values:

$$\begin{array}{c|c|c|c|c} x & y & x-\bar{x} & y-\bar{y} & (x-\bar{x})(y-\bar{y}) \\\hline 1 & 8 & -1.5 & 3 & -4.5 \\ 2 & 6 & -0.5 & 1 & -0.5 \\ 3 & 4 & 0.5 & -1 & -0.5 \\ 4 & 2 & 1.5 & -3 & -4.5 \end{array}$$

Then:

$$\sum (x-\bar{x})(y-\bar{y}) = -10$$ $$\sum (x-\bar{x})^2 = 5, \qquad \sum (y-\bar{y})^2 = 20$$

So:

$$r = \frac{-10}{\sqrt{5 \cdot 20}} = \frac{-10}{10} = -1$$

Interpretation: There is a perfect negative linear correlation.

Worked Example 3: A realistic data set

A teacher records study time and quiz score for five students.

• Study hours x: 1, 2, 3, 4, 5
• Quiz score y: 52, 55, 61, 65, 67

We expect a positive correlation, but probably not a perfect one.

First find the means:

$$\bar{x} = \frac{1+2+3+4+5}{5} = 3$$ $$\bar{y} = \frac{52+55+61+65+67}{5} = 60$$

Now calculate the deviations and products:

$$\begin{array}{c|c|c|c|c|c|c} x & y & x-\bar{x} & y-\bar{y} & (x-\bar{x})(y-\bar{y}) & (x-\bar{x})^2 & (y-\bar{y})^2 \\\hline 1 & 52 & -2 & -8 & 16 & 4 & 64 \\ 2 & 55 & -1 & -5 & 5 & 1 & 25 \\ 3 & 61 & 0 & 1 & 0 & 0 & 1 \\ 4 & 65 & 1 & 5 & 5 & 1 & 25 \\ 5 & 67 & 2 & 7 & 14 & 4 & 49 \end{array}$$

Add the columns:

$$\sum (x-\bar{x})(y-\bar{y}) = 40$$ $$\sum (x-\bar{x})^2 = 10$$ $$\sum (y-\bar{y})^2 = 164$$

Substitute:

$$r = \frac{40}{\sqrt{10 \cdot 164}} = \frac{40}{\sqrt{1640}}$$

Now approximate:

$$\sqrt{1640} \approx 40.50$$ $$r \approx \frac{40}{40.50} \approx 0.99$$

Interpretation: This is a very strong positive linear correlation. Students who studied more tended to have higher quiz scores.

But this still does not prove that studying alone caused the score increase. Other factors, such as prior understanding, sleep, or difficulty level, may also matter.

Worked Example 4: Correlation does not mean causation

Imagine a town collects data and finds that on days with more ice cream sales, there are also more people at the beach.

This could produce a positive correlation between ice cream sales and beach attendance.

Does that mean buying ice cream causes people to go to the beach? Not necessarily.

A third factor, such as hot weather, may be influencing both variables:

• Hotter days lead to more people going to the beach.
• Hotter days also lead to more ice cream being sold.

This is why we say correlation does not imply causation.

7. Important ideas when interpreting r

• r only measures linear relationships. If the pattern is curved, r may be misleading.
• Outliers can strongly affect r. One unusual point can make the correlation much stronger or weaker.
• The sign shows direction. Positive means together; negative means opposite.
• The size shows strength. Values closer to 1 or -1 mean a stronger linear pattern.
• Units do not affect r. Whether height is measured in inches or centimeters, the correlation stays the same.

8. Correlation and scatter plots

It helps to connect r with what you would see on a scatter plot.

• If points rise from left to right and cluster closely around a line, r is positive and near 1.
• If points fall from left to right and cluster closely around a line, r is negative and near -1.
• If points are widely scattered with no clear straight-line pattern, r is near 0.

A correlation value should always be interpreted together with the graph and the context of the data.

9. Common mistakes to avoid

• Mistake 1: Thinking r = 0 means no relationship at all. It only means no linear relationship.
• Mistake 2: Thinking a high correlation proves one variable causes the other. It does not.
• Mistake 3: Ignoring outliers. A single unusual point can change the value of r a lot.
• Mistake 4: Forgetting the sign. For example, r = -0.85 is strong, not weak; it is strongly negative.

10. Quick interpretation guide

Here are some sample values and meanings:

• r = 0.92: strong positive linear correlation
• r = 0.48: moderate positive linear correlation
• r = -0.76: strong negative linear correlation
• r = -0.12: weak negative linear correlation
• r = 0.00: no linear correlation

11. When is Pearson correlation useful?

Pearson's r is useful when:

• you have two numerical variables,
• the data are paired,
• you want to describe the direction and strength of a linear relationship.

It is often used before finding a line of best fit or making predictions with linear regression.

Brief Summary

The Pearson correlation coefficient, written as r, measures the direction and strength of a linear relationship between two numerical variables.

Its value is always between 1 and 1. Positive values show that the variables tend to increase together, negative values show that one tends to decrease as the other increases, and values near 0 show little or no linear relationship.

When interpreting correlation, always remember three key ideas:

• Look at the sign for direction.
• Look at the absolute value for strength.
• Remember that correlation does not imply causation.

If you can calculate r, read its meaning, and avoid confusing correlation with cause, you have understood the main idea of Pearson's correlation coefficient.

Least Squares Regression Line is a line that models the relationship between two quantitative variables. It is used when we want to predict a value of one variable from another and when the data points show a roughly linear trend.

For example, we might want to use hours studied to predict a test score, or temperature to predict ice cream sales. In these situations, each data point is an ordered pair \\((x,y)\\), where \\(x\\) is the explanatory variable and \\(y\\) is the response variable.

The least squares regression line is often called the line of best fit. Its job is to get as close as possible to all the data points overall. But “closest” needs a precise meaning, and that is where the idea of least squares comes in.

The equation of the least squares regression line is usually written as

$$\hat{y} = a + bx$$

where:

• \\(\hat{y}\\) is the predicted value of \\(y\\),
• \\(a\\) is the y-intercept,
• \\(b\\) is the slope.

The slope tells us how much the predicted value of \\(y\\) changes when \\(x\\) increases by 1 unit. The intercept tells us the predicted value of \\(y\\) when \\(x=0\\).

Why is it called least squares? For each data point, the regression line gives a predicted value. The difference between the actual value and the predicted value is called a residual:

$$\text{residual} = y - \hat{y}$$

A positive residual means the actual point is above the line. A negative residual means the actual point is below the line.

The least squares regression line is the line that makes the sum of the squared residuals as small as possible:

$$\text{Residual Sum of Squares} = \sum (y-\hat{y})^2$$

We square the residuals so that positive and negative differences do not cancel out, and so larger errors count more heavily.

Key formulas for the regression line are:

$$b = r\left(\frac{s_y}{s_x}\right)$$ $$a = \bar{y} - b\bar{x}$$

where:

• \\(r\\) is the correlation coefficient,
• \\(s_x\\) is the standard deviation of the \\(x\\)-values,
• \\(s_y\\) is the standard deviation of the \\(y\\)-values,
• \\(\bar{x}\\) is the mean of the \\(x\\)-values,
• \\(\bar{y}\\) is the mean of the \\(y\\)-values.

Another important fact is that the least squares regression line always passes through the point \\((\bar{x},\bar{y})\\).

In many school problems, a calculator or software gives the regression line directly. Even then, you still need to understand what the slope and intercept mean and how to use the line for prediction.

How to interpret the regression line

• The slope should be interpreted in context: “For each increase of 1 unit in \\(x\\), the predicted \\(y\\) changes by about ... units.”
• The intercept should also be interpreted in context, but only if \\(x=0\\) makes sense in the situation.
• The line can be used to make predictions by substituting a value of \\(x\\) into the equation.

Important caution: A regression line is most reliable for values of \\(x\\) that are within the range of the original data. Predicting far outside the data range is called extrapolation, and it can be misleading.

Worked Example 1: Understanding slope and intercept

A regression line relating hours studied \\(x\\) and test score \\(y\\) is

$$\hat{y} = 52 + 6x$$

Interpret the slope and intercept, and predict the score for a student who studies 4 hours.

Step 1: Interpret the slope.

The slope is 6. This means that for each additional hour studied, the predicted test score increases by about 6 points.

Step 2: Interpret the intercept.

The intercept is 52. This means that when a student studies 0 hours, the predicted test score is 52.

Step 3: Make a prediction.

Substitute \\(x=4\\) into the equation:

$$\hat{y} = 52 + 6(4) = 52 + 24 = 76$$

Answer: The predicted score for a student who studies 4 hours is 76.

Worked Example 2: Using summary statistics to find the regression line

Suppose for a set of data we know:

• \\(\bar{x} = 10\\)
• \\(\bar{y} = 25\\)
• \\(s_x = 2\\)
• \\(s_y = 5\\)
• \\(r = 0.8\\)

Find the least squares regression line.

Step 1: Find the slope.

$$b = r\left(\frac{s_y}{s_x}\right) = 0.8\left(\frac{5}{2}\right) = 0.8(2.5) = 2$$

Step 2: Find the intercept.

$$a = \bar{y} - b\bar{x} = 25 - 2(10) = 25 - 20 = 5$$

Step 3: Write the equation.

$$\hat{y} = 5 + 2x$$

Answer: The least squares regression line is

$$\hat{y} = 5 + 2x$$

Notice that this line passes through \\((10,25)\\), since

$$5 + 2(10) = 25$$

Worked Example 3: Residuals and what they mean

A regression line predicting sales \\(y\\) from advertising spending \\(x\\) is

$$\hat{y} = 120 + 8x$$

If a business spends \(\$5\\) thousand on advertising, the actual sales are \(\$170\\) thousand. Find the predicted sales and the residual.

Step 1: Find the predicted sales.

$$\hat{y} = 120 + 8(5) = 120 + 40 = 160$$

So the predicted sales are \(\$160\\) thousand.

Step 2: Find the residual.

$$\text{residual} = y - \hat{y} = 170 - 160 = 10$$

Answer: The residual is 10. Since it is positive, the actual sales were \(\$10\\) thousand higher than predicted by the line.

Worked Example 4: A full interpretation in context

A school records the relationship between number of absences \\(x\\) and final grade \\(y\\). The least squares regression line is

$$\hat{y} = 91 - 2.5x$$

Answer the following:

1. Interpret the slope.
2. Interpret the intercept.
3. Predict the grade for a student with 6 absences.
4. Decide whether predicting the grade for 40 absences is reasonable if the original data only included students with 0 to 12 absences.

Step 1: Interpret the slope.

The slope is \\(-2.5\\). For each additional absence, the predicted final grade decreases by about 2.5 points.

Step 2: Interpret the intercept.

The intercept is 91. This means that a student with 0 absences is predicted to have a final grade of 91.

Step 3: Predict the grade for 6 absences.

$$\hat{y} = 91 - 2.5(6) = 91 - 15 = 76$$

The predicted grade is 76.

Step 4: Consider extrapolation.

Predicting for 40 absences is not reasonable because 40 is far outside the data range of 0 to 12 absences. This would be extrapolation, and the linear model may no longer be accurate.

When is a least squares regression line appropriate?

• When both variables are quantitative.
• When the scatter plot shows a roughly linear pattern.
• When there are no extreme outliers that completely distort the trend.

When should you be careful?

• If the relationship is curved rather than linear.
• If there are outliers that strongly affect the line.
• If you are interpreting an intercept that has no real meaning in context.
• If you are extrapolating beyond the observed data.

Common mistakes to avoid

• Confusing the actual value \\(y\\) with the predicted value \\(\hat{y}\\).
• Forgetting that residual is \\(y-\hat{y}\\), not \\(\hat{y}-y\\).
• Giving a slope interpretation without units or context.
• Treating correlation as proof that one variable causes the other.
• Using the model for values far outside the data set.

Step-by-step method for regression questions

1. Identify the explanatory variable \\(x\\) and response variable \\(y\\).
2. Write the regression equation in the form \\(\hat{y} = a + bx\\).
3. Interpret the slope in context.
4. Interpret the intercept in context, if it makes sense.
5. Substitute a given \\(x\\)-value to make a prediction.
6. If asked, calculate the residual using \\(y-\hat{y}\\).
7. Check whether the prediction is interpolation or extrapolation.

Brief Summary

The least squares regression line is a linear model that predicts \\(y\\) from \\(x\\) while minimizing the sum of squared residuals. Its equation is \\(\hat{y}=a+bx\\), where the slope describes how predicted \\(y\\) changes as \\(x\\) changes, and the intercept gives the predicted value when \\(x=0\\). It is useful for making predictions and describing trends, but it should be used carefully, especially when extrapolating or when the data are not linear.

Introduction to Confidence Intervals and Hypothesis Testing

In statistics, we often want to learn about a whole population, but we usually only have data from a sample. For example, we may want to know the average height of all students in a school, or whether a new study method really improves test scores.

Because a sample is only part of the population, our results are not exact. Two important tools help us make sensible conclusions from sample data: confidence intervals and hypothesis tests.

A confidence interval gives a range of likely values for a population parameter, such as a population mean or proportion. A hypothesis test uses sample evidence to decide whether a claim about a population seems reasonable or not.

These ideas are closely connected. A confidence interval estimates a parameter, while a hypothesis test checks a claim about that parameter.

1. Population, Sample, Parameter, and Statistic

Before learning the methods, it is important to know some key vocabulary.

• Population: the entire group we want to study
• Sample: a smaller group taken from the population
• Parameter: a number that describes the population, such as the population mean \(\mu\) or population proportion \(p\)
• Statistic: a number calculated from the sample, such as the sample mean \(\bar{x}\) or sample proportion \(\hat{p}\)

Since parameters are usually unknown, we use statistics to estimate them.

2. What Is a Confidence Interval?

A confidence interval is an interval estimate for a population parameter. It usually has the form:

$$\text{estimate} \pm \text{margin of error}$$

For a population mean, the estimate is often the sample mean \(\bar{x}\). For a population proportion, the estimate is often the sample proportion \(\hat{p}\).

The margin of error tells us how much uncertainty there is in the estimate. A smaller margin of error means a more precise estimate.

For a simple introduction, we can think of a confidence interval as:

$$\text{sample statistic} \pm \text{critical value} \times \text{standard error}$$

You do not need every detail right away. The main idea is that sample results vary, so we give a range of believable values instead of one exact answer.

3. Interpreting a Confidence Level

You will often hear about a 95% confidence interval. This does not mean there is a 95% chance that the true population value is inside this one specific interval.

Instead, it means that if we repeatedly took many random samples and built a confidence interval from each one, about 95% of those intervals would contain the true population parameter.

In 12th Grade terms, a 95% confidence interval is a range produced by a method that works correctly about 95% of the time.

4. Confidence Interval for a Population Proportion

Suppose we survey \(n\) people and \(x\) of them give a certain response. Then the sample proportion is

$$\hat{p} = \frac{x}{n}$$

A commonly used confidence interval for a population proportion is

$$\hat{p} \pm z^*\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$$

Here, \(z^*\) is a critical value based on the confidence level. For a 95% confidence level, \(z^* \approx 1.96\).

5. Confidence Interval for a Population Mean

If we are estimating a population mean from a sample with mean \(\bar{x}\), then a common confidence interval is

$$\bar{x} \pm z^*\left(\frac{\sigma}{\sqrt{n}}\right)$$

Here, \(\sigma\) is the population standard deviation if it is known. In many school-level introductions, this formula is used to understand the basic idea of confidence intervals.

The standard error becomes smaller when the sample size \(n\) gets larger, because \(\sqrt{n}\) is in the denominator. This means larger samples usually produce narrower confidence intervals.

6. What Affects the Width of a Confidence Interval?

• Sample size: Larger samples give narrower intervals.
• Confidence level: Higher confidence levels give wider intervals.
• Variability: More spread in the data gives wider intervals.

There is a trade-off between confidence and precision. If you want to be more confident, you usually need a wider interval.

Worked Example 1: Confidence Interval for a Proportion

A school surveys 200 students about whether they support a new lunch option. Out of 200 students, 124 say yes. Find an approximate 95% confidence interval for the true proportion of students who support the new lunch option.

Step 1: Find the sample proportion.

$$\hat{p} = \frac{124}{200} = 0.62$$

Step 2: Use the formula.

$$\hat{p} \pm 1.96\sqrt{\frac{\hat{p}(1-\hat{p})}{n}}$$ $$0.62 \pm 1.96\sqrt{\frac{0.62(0.38)}{200}}$$

Step 3: Calculate the standard error.

$$\sqrt{\frac{0.62 \times 0.38}{200}} = \sqrt{\frac{0.2356}{200}} = \sqrt{0.001178} \approx 0.0343$$

Step 4: Find the margin of error.

$$1.96 \times 0.0343 \approx 0.0672$$

Step 5: Write the interval.

$$0.62 \pm 0.0672$$ $$[0.5528, 0.6872]$$

Interpretation: We are 95% confident that the true proportion of all students who support the lunch option is between about 55.3% and 68.7%.

7. What Is Hypothesis Testing?

A hypothesis test is a method for checking whether sample data supports or contradicts a claim about a population.

Every hypothesis test begins with two hypotheses:

• Null hypothesis \((H_0)\): the original claim or a statement of no change, no effect, or no difference
• Alternative hypothesis \((H_a)\): the claim we want evidence for

For example, if a company says the average battery life is 10 hours, then

$$H_0: \mu = 10$$ $$H_a: \mu \neq 10$$

The null hypothesis uses equality. The alternative can be:

• \(H_a: \mu \neq \text{value}\) for “different from”
• \(H_a: \mu > \text{value}\) for “greater than”
• \(H_a: \mu < \text{value}\) for “less than”

8. The Logic of a Hypothesis Test

The idea is simple: assume the null hypothesis is true, and then ask whether the sample result would be unusual under that assumption.

If the sample result is not unusual, we fail to reject \(H_0\). If the sample result is very unusual, we reject \(H_0\).

Rejecting \(H_0\) means the sample gives evidence in favor of the alternative hypothesis. Failing to reject \(H_0\) does not prove that \(H_0\) is true. It only means there is not enough evidence against it.

9. Test Statistic and p-value

In many tests, we convert the sample result into a test statistic, often a z-score:

$$z = \frac{\text{sample result} - \text{hypothesized value}}{\text{standard error}}$$

This tells us how far the sample result is from what the null hypothesis predicts, measured in standard errors.

Then we find the p-value. The p-value is the probability, assuming \(H_0\) is true, of getting a result at least as extreme as the one observed.

Small p-value: strong evidence against \(H_0\)

Large p-value: weak evidence against \(H_0\)

A common cutoff is the significance level \(\alpha = 0.05\).

• If \(p\text{-value} \leq 0.05\), reject \(H_0\).
• If \(p\text{-value} > 0.05\), fail to reject \(H_0\).

10. Hypothesis Test for a Population Proportion

If the null hypothesis claims \(p = p_0\), then for a one-sample z-test for a proportion, we use

$$z = \frac{\hat{p}-p_0}{\sqrt{\frac{p_0(1-p_0)}{n}}}$$

Notice that the standard error uses \(p_0\), the proportion from the null hypothesis.

Worked Example 2: Hypothesis Test for a Proportion

A website claims that 70% of users can complete a task in under 3 minutes. A teacher tests this with 100 students, and only 62 complete it in under 3 minutes. Test the claim at the 5% significance level.

Step 1: State the hypotheses.

$$H_0: p = 0.70$$ $$H_a: p < 0.70$$

This is a left-tailed test because we are checking if the true proportion is less than 70%.

Step 2: Find the sample proportion.

$$\hat{p} = \frac{62}{100} = 0.62$$

Step 3: Compute the test statistic.

$$z = \frac{0.62-0.70}{\sqrt{\frac{0.70(0.30)}{100}}}$$ $$z = \frac{-0.08}{\sqrt{0.0021}}$$ $$z = \frac{-0.08}{0.0458} \approx -1.75$$

Step 4: Find the p-value.

A z-score of \(-1.75\) gives a left-tail p-value of about \(0.040\).

Step 5: Make a decision.

Since \(0.040 < 0.05\), we reject \(H_0\).

Conclusion: There is enough evidence at the 5% level to suggest that fewer than 70% of users can complete the task in under 3 minutes.

11. Hypothesis Test for a Population Mean

If a null hypothesis says \(\mu = \mu_0\), then an introductory z-test for a mean uses

$$z = \frac{\bar{x}-\mu_0}{\sigma/\sqrt{n}}$$

This compares the sample mean to the hypothesized population mean.

Worked Example 3: Hypothesis Test for a Mean

A factory states that the mean mass of a packet is 500 g. Suppose the population standard deviation is known to be 20 g. A sample of 64 packets has mean mass 495 g. Test at the 5% significance level whether the true mean mass is different from 500 g.

Step 1: State the hypotheses.

$$H_0: \mu = 500$$ $$H_a: \mu \neq 500$$

This is a two-tailed test because we are checking for any difference.

Step 2: Compute the test statistic.

$$z = \frac{495-500}{20/\sqrt{64}}$$ $$z = \frac{-5}{20/8} = \frac{-5}{2.5} = -2$$

Step 3: Find the p-value.

For \(z=-2\) in a two-tailed test,

$$p\text{-value} \approx 2(0.0228) = 0.0456$$

Step 4: Make a decision.

Since \(0.0456 < 0.05\), we reject \(H_0\).

Conclusion: There is enough evidence to suggest that the true mean mass is different from 500 g.

12. Connection Between Confidence Intervals and Hypothesis Tests

Confidence intervals and hypothesis tests are strongly linked.

For a 95% confidence interval and a 5% hypothesis test:

• If the hypothesized value is inside the 95% confidence interval, then we usually fail to reject \(H_0\) at the 5% level.
• If the hypothesized value is outside the 95% confidence interval, then we usually reject \(H_0\) at the 5% level.

This gives a useful way to check whether a claim fits the sample data.

Worked Example 4: Using a Confidence Interval to Judge a Claim

A random sample estimates the mean time students spend on homework each night. The sample gives a 95% confidence interval of \([1.8, 2.4]\) hours. A school report claims that the population mean is 2.6 hours. Is this claim supported at the 5% significance level?

Since \(2.6\) is not in the 95% confidence interval \([1.8, 2.4]\), the claim is not supported by the sample at the 5% significance level.

So we would reject the null hypothesis \(H_0: \mu = 2.6\).

13. Type I and Type II Errors

Because we make decisions using sample data, mistakes are possible.

• Type I error: rejecting \(H_0\) when it is actually true
• Type II error: failing to reject \(H_0\) when it is actually false

If we choose a significance level of \(\alpha = 0.05\), we are accepting a 5% chance of a Type I error.

At this level, if the null hypothesis is really true, there is a 5% chance that our test will still wrongly reject it.

14. Important Conditions and Cautions

Confidence intervals and hypothesis tests work best when the sample is random and the data meet suitable conditions.

• The sample should represent the population fairly.
• A larger sample usually gives more reliable results.
• Statistical significance does not always mean practical importance.

For example, a tiny difference may be statistically significant if the sample size is very large, but it may not matter much in real life.

15. Common Mistakes to Avoid

• Thinking that a 95% confidence interval means a 95% chance the parameter is in that exact interval
• Saying “accept \(H_0\)” instead of “fail to reject \(H_0\)”
• Forgetting whether the test is left-tailed, right-tailed, or two-tailed
• Mixing up the sample statistic with the population parameter
• Ignoring the context when interpreting results

16. Step-by-Step Method

For a confidence interval:

1. Identify the parameter to estimate.
2. Compute the sample statistic.
3. Find the standard error.
4. Choose the confidence level and critical value.
5. Calculate the margin of error.
6. Write and interpret the interval in context.

For a hypothesis test:

1. State \(H_0\) and \(H_a\).
2. Choose the significance level \(\alpha\).
3. Calculate the sample statistic and test statistic.
4. Find the p-value.
5. Reject or fail to reject \(H_0\).
6. Write a conclusion in context.

Brief Summary

Confidence intervals and hypothesis tests are tools for drawing conclusions about a population from a sample.

A confidence interval gives a range of likely values for a population parameter, while a hypothesis test checks whether a specific claim about that parameter is supported by the data.

Both methods rely on understanding sampling variation. The bigger idea is that sample results naturally change, so statistics helps us decide what conclusions are reasonable.