STAT 2000 exam review
A free review page for UGA’s STAT 2000: the units in order, eight worked examples that spend their words on interpretation, and ten practice questions with answers. Nothing to buy and nothing to sign up for.
By Michael Rainwater. Read it alongside how to pass STAT 2000, which explains why interpretation is where the points go.
Finals Exam Intensive. UGA’s fall finals run December 3 to 9, 2026; your section’s date is on the Registrar’s final exam schedule. If you want the last week built around what will cost you points, the STAT 2000 Exam Intensive starts with a diagnostic exam and runs on Zoom or in person in Athens.
What the exam asks you to do, unit by unit
The units below follow the order in my STAT 2000 guide. Your instructor may reorder them, so use your syllabus for what your exam covers. The right column is the part that matters: on this course the exam asks why, not which formula.
| Unit | What the exam asks you to do |
|---|---|
| Sampling and study design | Say what a conclusion can claim: random sampling lets you generalize, random assignment lets you claim cause. |
| Numerical and graphical summaries | Choose the right center and spread for the shape, and spot an outlier. |
| Correlation and regression | Interpret a slope and an r² in context, and refuse to extrapolate. |
| Contingency tables | Read a conditional probability from the right row or column, and say whether two variables look related. |
| Probability and the normal distribution | Standardize, shade the region, and put the answer in a sentence. |
| Sampling distributions | Explain why the sample mean varies less than the data, and what the central limit theorem lets you assume. |
| Confidence intervals | Build the interval, then read it correctly. This is where wrong readings cost the most. |
| Hypothesis tests | Name the parameter, check the conditions, run the test, and state the conclusion in context. |
Worked examples
Each one ends with the sentence I would want to see on your exam paper. Critical values are stated when they are used. For every inference problem I follow the same order: parameter, test, conditions, calculation, conclusion in context.
Example 1: Center and spread for a skewed sample
Nine students report how many hours they worked last week: 12, 14, 15, 15, 16, 18, 19, 21, 45. Which summary should you report, and why?
Solution. The mean is 19.4 hours and the median is 16 hours. The standard deviation is 9.96 hours. The 45 pulls the mean up by about 3 hours and inflates the standard deviation. Without it, the mean is 16.25 and the standard deviation is 2.92. Using the median of each half for the quartiles, Q1 = 14.5 and Q3 = 20, so the IQR is 5.5. The upper fence is 20 + 1.5(5.5) = 28.25, and 45 is far past it, so it is an outlier by the 1.5 IQR rule. (Software may compute quartiles slightly differently; the conclusion does not change.)
The sentence: the distribution is skewed right with one outlier, so the median (16 hours) and IQR (5.5 hours) describe the typical student better than the mean and standard deviation, which the outlier distorts.
Example 2: Interpreting a regression line
For a class, hours studied per week has mean 6 and standard deviation 2.5. Exam score has mean 72 and standard deviation 9. The correlation is r = 0.80. Find the line, interpret it, and say whether to predict for a student who studies 20 hours.
Solution. Slope: b = r(sy/sx) = 0.80(9/2.5) = 2.88. Intercept: a = 72 − 2.88(6) = 54.72. The line is predicted score = 54.72 + 2.88(hours). At 8 hours the prediction is 77.76. The coefficient of determination is r² = 0.64.
The sentences: for each additional hour studied per week, the predicted exam score rises by 2.88 points, for students like these. About 64% of the variation in exam scores is explained by the linear relationship with hours studied. The intercept, 54.72, is the predicted score at zero hours, which is only meaningful if zero hours is within the data. At 20 hours the line gives 112.32, a score above 100. That is extrapolation, far outside the data, and the line should not be trusted there. Also, a strong correlation does not show that studying causes the higher scores.
Example 3: A normal distribution, in both directions
Scores on a test are approximately normal with mean 78 and standard deviation 8. (a) What proportion score above 90? (b) What score marks the 90th percentile?
Solution. (a) z = (90 − 78)/8 = 1.50. The area above z = 1.50 is 0.0668. (b) The z-score with 90% of the area below it is 1.2816, so the score is 78 + 1.2816(8) = 88.25.
The sentences: a 90 is 1.5 standard deviations above the mean, and about 6.7% of students score higher. A score of about 88 is the 90th percentile, meaning about 90% of scores fall at or below it. The empirical rule agrees with the picture: about 68% of scores (0.6827 exactly) fall within one standard deviation, between 70 and 86.
Example 4: Why an observational study cannot say “because”
Students who attend optional review sessions average 8 points higher on the exam than students who do not. Can you conclude that the sessions raise scores?
Solution. No. Students chose whether to attend, so this is an observational study. Students who attend may also be more motivated, which affects both attendance and score. That is a confounding variable, and the study cannot separate its effect from the effect of the sessions. To claim cause, randomly assign students to attend or not, which makes the groups alike on average in everything else, motivation included.
The sentence: the sessions are associated with higher scores, but because students were not randomly assigned, a confounder such as motivation could explain the difference. Random sampling and random assignment answer different questions: sampling randomly lets you generalize to a population, and assigning randomly lets you claim cause.
Example 5: Conditional probability from a table
Of 300 students surveyed, 150 are underclassmen and 150 are upperclassmen. Among the underclassmen, 60 have a job and 90 do not. Among the upperclassmen, 105 have a job and 45 do not. Find P(job), P(job given upperclass), and P(upperclass given job). Are class standing and having a job independent?
Solution. 165 of the 300 have a job, so P(job) = 165/300 = 0.55. Restricting to the upperclassmen row, P(job given upperclass) = 105/150 = 0.70. Restricting to the job column, P(upperclass given job) = 105/165 = 0.636. For independence, P(job given upperclass) would equal P(job). Here 0.70 is not 0.55 (and underclassmen are at 60/150 = 0.40), so they are not independent.
The sentence: 70% of upperclassmen have a job, but 63.6% of students with a job are upperclassmen. These are different questions with different answers, and the one you want depends on which group the problem says to start from. Having a job is more common among upperclassmen than overall, so the two variables are associated.
Example 6: The sampling distribution and the central limit theorem
A population has mean 100 and standard deviation 12, and is not normal. You take a random sample of n = 36. Describe the distribution of the sample mean, then find P(sample mean > 103).
Solution. The mean of the sample mean is 100. Its standard deviation (the standard error) is 12/√36 = 2. By the central limit theorem, a sample of 36 is generally large enough for the distribution of the sample mean to be approximately normal even though the population is not. Then z = (103 − 100)/2 = 1.50, and P(sample mean > 103) = 0.0668. For comparison, one individual value above 103 has z = 0.25 and probability 0.4013.
The sentence: one student scoring above 103 is common (about 40%), but a sample average above 103 is unusual (about 6.7%), because averages of 36 values vary far less than single values do. The standard error describes how much the sample mean varies from sample to sample, not how much the data vary.
Example 7: A confidence interval for a proportion, read correctly
In a random sample of 600 UGA students, 312 say they prefer a morning exam time. Build and interpret a 95% confidence interval for the population proportion.
Solution. Parameter: p, the proportion of all UGA students who prefer morning. Conditions: random sample, and both 312 and 288 are at least 10, so the normal approximation is reasonable. p-hat = 312/600 = 0.52. Standard error = √(0.52 × 0.48 / 600) = 0.0204. For 95% confidence use z* = 1.96, so the margin of error is 1.96(0.0204) = 0.040. The interval is 0.52 ± 0.040, or (0.480, 0.560).
The sentence: I am 95% confident that the proportion of all UGA students who prefer a morning exam time is between 48.0% and 56.0%. The 95% describes the method: about 95% of intervals built this way capture the true proportion. It does not say that 95% of students fall in this range. And because the interval includes 0.50, these data do not show that a majority prefer mornings.
Example 8: A confidence interval and tests for means
(a) A t interval. A random sample of 25 students sleeps an average of 6.8 hours with a sample standard deviation of 1.5 hours. The population standard deviation is unknown, so use a t interval with df = 24. The critical value is t* = 2.064 for 95%. The standard error is 1.5/√25 = 0.30, so the margin of error is 2.064(0.30) = 0.619 and the interval is (6.18, 7.42) hours. You would first check that the sample is random and that the sample is not strongly skewed. I am 95% confident that the mean nightly sleep of all students in this population is between 6.18 and 7.42 hours.
(b) A one-sample t test. A course claims the mean score is 70. A random sample of 16 students has mean 73.2 and standard deviation 8. Test H0: μ = 70 against Ha: μ ≠ 70. The standard error is 8/√16 = 2, so t = (73.2 − 70)/2 = 1.60 with df = 15, and the two-sided p-value is 0.130. If the true mean were 70, a sample mean at least this far from 70 would happen about 13% of the time just by chance. At the 0.05 level I fail to reject H0: the data do not provide convincing evidence that the mean differs from 70. That is not the same as showing the mean is 70.
(c) A two-sample t test. In a randomized experiment, 40 students study with spaced practice and average 78.5 (standard deviation 9), and 35 students study the same material in one block and average 73.1 (standard deviation 10). Test H0: μ1 − μ2 = 0 against a two-sided alternative. The standard error is √(9²/40 + 10²/35) = 2.21, so t = 5.4/2.21 = 2.44. Software gives df of about 69 and a two-sided p-value of 0.017. A 95% interval for the difference is 5.4 ± 1.995(2.21), or (0.99, 9.81) points. Because the groups were randomly assigned, the conclusion can be causal: there is convincing evidence at the 0.05 level that spaced practice raises the mean score compared with one block, and I am 95% confident the mean difference is between about 1 and 10 points. The interval does not contain 0, which agrees with rejecting H0. It also shows the effect could be small.
Ten practice questions
Work each one on paper first, in full sentences where the question asks for a conclusion. Then open the answer. Every answer here was checked by computation.
- Scores have mean 70 and standard deviation 12. What is the z-score of a 91, and what does it mean?
Answer
z = (91 − 70)/12 = 1.75. The score is 1.75 standard deviations above the mean.
- A distribution is skewed right. Which is larger, the mean or the median, and why?
Answer
The mean. The long right tail pulls the mean up, while the median resists extreme values.
- A 95% confidence interval for the mean hours of sleep is (6.6, 7.0). A friend says 95% of students sleep between 6.6 and 7.0 hours. What is wrong?
Answer
The interval is about the population mean, not about individual students. Individual students vary much more than the mean does. The 95% describes how often the method captures the true mean.
- You quadruple the sample size. What happens to the standard error of the sample mean?
Answer
It is cut in half, because the standard error is σ/√n and √4 = 2.
- A test gives a p-value of 0.03. What is the decision at the 0.05 level, and at the 0.01 level?
Answer
At 0.05, reject H0 because 0.03 < 0.05. At 0.01, fail to reject because 0.03 > 0.01. The p-value is the probability of data this extreme if H0 is true; it is not the probability that H0 is true.
- A regression line is predicted score = 20 + 3(hours). Predict the score at 5 hours, interpret the slope, and find the residual for a student at 5 hours who scored 41.
Answer
Prediction = 20 + 3(5) = 35. Each extra hour is associated with 3 more predicted points. Residual = observed − predicted = 41 − 35 = +6, so this student scored 6 points above the line.
- The same 30 students take a quiz before and after a review session. Is this a one-sample, two-sample, or paired design, and why?
Answer
Paired. Each student gives two measurements, so the analysis works with the 30 differences and not two independent groups.
- Is a 99% confidence interval wider or narrower than a 95% interval from the same data? Give the critical values for a z interval.
Answer
Wider. z* = 2.576 for 99% against z* = 1.96 for 95%. More confidence costs a wider interval.
- A school asks students to text in their opinion on a new policy, and 72% of those who respond are in favor. Why can you not say 72% of all students agree?
Answer
It is a voluntary response sample. People with strong opinions choose to respond, so the sample is biased and the result cannot be generalized to the population. A random sample would be needed.
- In a random sample of 150 students, 45 own a bike. Build a 95% confidence interval for the population proportion and interpret it.
Answer
p-hat = 45/150 = 0.30. Standard error = √(0.30 × 0.70/150) = 0.0374. With z* = 1.96, the margin of error is 0.0733 and the interval is (0.227, 0.373). I am 95% confident that between 22.7% and 37.3% of all students in this population own a bike. The conditions hold: random sample, 45 and 105 are both at least 10.
How to use this in the last week
Do not reread. Cover the answer and write it out.
- Three or four days out: do the practice questions and the examples cold, then sort each miss into the four types from the guide: wrong test, condition ignored, calculation slip, or interpretation. Most of your misses will be the last one.
- Two days out: take a full practice exam from your own course materials, timed, and write the conclusion sentence for every inference problem. If your sampling distribution explanation still sounds memorized, work that unit again before anything else.
- The day before: do one mixed set, then stop. Check your syllabus or instructor for what you may bring and what you must know.
Questions I get
Anything not covered here, ask me directly. (706) 389-0528.
Want help getting ready for the exam?
I tutor STAT 2000 one-on-one, over Zoom or in person in Athens. Tell me the exam date and what you got on the last one, and I will tell you what I would do with the time you have.