How to give each student a follow-up quiz based on what they missed
Two students both score 70% on a statistics quiz. One misses p-values, the other confidence intervals. Here is the follow-up quiz each one gets, and how to make one for every student in your Canvas course.
Two students in an intro statistics course both score 70% on Quiz 4. Same grade, different problem. Priya missed all three questions on p-values. Marcus missed all three on confidence intervals.
A review session gives them both the same practice, and half of it is on what each one already knows. This post shows the follow-up quiz each of them gets instead, and how to give one to every student in a Canvas course without writing any of them yourself.
1. Look at what each student missed
Here are the questions each one got wrong, and the answer they chose. Both got the other seven right. The students and their answers are made up for this post.
Priya: 7 of 10
- Question 1. A test of gives . Which statement is correct?Chose There is a 3% chance that the null hypothesis is true.Right If were 50, a sample mean at least this far from 50 would happen about 3% of the time.
- Question 2. A test at gives . What should you conclude?Chose Accept : the data show that .Right Fail to reject : the data do not give strong evidence against it.
- Question 3. Study A finds and Study B finds for the same claim. What can you conclude?Chose Study A found a larger effect than Study B.Right Study A's data give stronger evidence against .
Marcus: 7 of 10
- Question 4. A 95% confidence interval for the mean commute time is (22.1, 27.9) minutes. Which interpretation is correct?Chose 95% of commuters take between 22.1 and 27.9 minutes.Right We are 95% confident that the population mean commute time is between 22.1 and 27.9 minutes.
- Question 5. With the same sample, you raise the confidence level from 95% to 99%. What happens to the interval?Chose It gets narrower.Right It gets wider.
- Question 6. A sample of has mean 25 and standard deviation 6. What is the margin of error of a 95% confidence interval for the mean? Use .Chose 11.76Right 1.96
The wrong choices tell you more than the score. Priya reads a p-value as the chance that the null hypothesis is true, and a large one as proof that it is. Marcus reads a confidence interval as the range of individual values, and his 11.76 is the margin of error without dividing by . Each needs practice on their own idea, asked a new way, so a remembered answer does not fit.
2. The follow-up quiz each one gets
We gave the LMSystems question writer the six misses above, in the same form it reads them from a Canvas quiz, and asked for five questions each, short answer and multiple choice. It wrote the questions below on November 22, 2026. We did not change them.
Priya's follow-up quiz
A two-sided test of produces a p-value of 0.08. Which interpretation is correct?
- There is an 8% probability that the population mean equals 120.
- Assuming the null hypothesis and test assumptions hold, results at least as extreme as those observed occur about 8% of the time.Right answer
- There is a 92% probability that the alternative hypothesis is true.
- The sample mean is within 8% of the population mean.
Why Missed question 1 of Quiz 4: practices interpreting a p-value conditionally rather than as the probability that the null is true.
A researcher tests whether the mean battery life is 10 hours and obtains at . State the decision and explain whether the result establishes that the mean is 10 hours.
Answer key Fail to reject the null hypothesis because 0.27 exceeds 0.05. The data do not provide sufficient evidence that the mean differs from 10 hours, but this does not establish that the mean equals 10 hours.
Why Missed question 2 of Quiz 4: practices distinguishing failure to reject a null hypothesis from proving it true.
Two studies test the same null hypothesis using comparable methods: Study C reports and Study D reports . Which gives stronger evidence against the null, and can you determine which has the larger estimated effect from these p-values alone?
Answer key Study D gives stronger evidence against the null because its p-value is smaller. The p-values alone do not establish which estimated effect is larger, because p-values also depend on precision, including sample size and variability.
Why Missed question 3 of Quiz 4: practices separating strength of evidence from estimated effect size.
A simulation generates 2,000 samples under a null hypothesis; 90 produce test statistics at least as extreme as the observed statistic. What is the estimated p-value?
- 0.955
- 0.090
- 0.045Right answer
- 0.450
Why Missed question 1 of Quiz 4: practices obtaining a p-value from the proportion of extreme results under the null.
Two independent studies each estimate a treatment–control mean difference of 4 points and have similar within-group variability. Study E has 40 participants per group, while Study F has 400 per group; why would Study F generally have a smaller p-value for testing a zero difference?
- Study F has a larger estimated treatment effect.
- Study F generally has a smaller standard error, making the same estimated difference more extreme relative to its uncertainty.Right answer
- A larger sample makes the null hypothesis's probability smaller by definition.
- The significance level automatically decreases as sample size increases.
Why Missed question 3 of Quiz 4: practices explaining how sample size can change a p-value without changing the estimated effect.
Marcus's follow-up quiz
A 90% confidence interval for mean battery life is (11.2, 12.6) hours. A manager concludes that 90% of individual batteries last between these limits. What is wrong with this conclusion?
- The interval estimates the population mean, not the range containing 90% of individual battery lifetimes.Right answer
- The interval contains exactly 90% of the battery lifetimes in the sample, not the population.
- The manager should say that 10% of individual batteries last between these limits.
- A confidence interval describes only the sample mean and provides no information about the population mean.
Why Missed confidence-interval interpretation in Quiz 4: distinguishes estimation of a population mean from the spread of individual observations.
A researcher computes 90%, 95%, and 99% confidence intervals for a population mean using the same sample and method. Which ordering of their margins of error is correct?
- 99% < 95% < 90%
- 90% < 95% < 99%Right answer
- All three margins of error are equal.
- The ordering depends on the sample mean.
Why Missed confidence-level question in Quiz 4 (said higher confidence gives a narrower interval): practices the relationship between confidence and margin of error.
A random sample of 100 deliveries has a mean delivery time of 42 minutes and a standard deviation of 15 minutes. Using , calculate the margin of error for a 99% confidence interval for the population mean; show how sample size enters your calculation.
Answer key The estimated standard error is minutes. The margin of error is minutes, or approximately 3.86 minutes.
Why Missed margin-of-error calculation in Quiz 4 (gave 11.76): practices multiplying the critical value by the standard error rather than the sample standard deviation.
Researchers repeatedly draw random samples from one population and construct a 95% confidence interval for its mean from each sample. Over many repetitions, what should happen about 95% of the time, and what stays fixed across the repetitions?
Answer key About 95% of the intervals should contain the true population mean, assuming the method's conditions hold. The population mean stays fixed, while the samples and interval endpoints vary.
Why Missed confidence-interval interpretation in Quiz 4: practices the repeated-sampling meaning of confidence.
Two samples have the same standard deviation and use the same critical value to estimate a population mean. If sample size increases from 25 to 100, what happens to the margin of error?
- It stays the same because the standard deviation is unchanged.
- It becomes one-fourth as large.
- It becomes twice as large.
- It becomes half as large.Right answer
Why Missed margin-of-error calculation in Quiz 4: reinforces the square-root relationship between sample size, standard error, and margin of error.
Neither student sees a quiz question again. Priya's first question offers her own wrong idea as a choice, to test if it is gone. Marcus has to show where the sample size goes in a margin of error. Under each question, Why names the miss it practices. Only you see it, never the student.
3. Make one for every student in your course
In LMSystems this is a personal quiz. You make one for a whole Canvas course at once:
- Connect Canvas. Sign in with Canvas if your school is set up for it, or paste an access token you make in your Canvas account settings.
- Pick a course. Choose how many questions each student gets, 5 to 10, and of which kinds. Then press Write each student's questions.
- It reads what each student missed. Quiz questions they got wrong or partly right, rubric criteria where they lost points, low scores, missing work, and answers under full marks on the course's LMSystems quizzes. Then it writes each student's set.
- Check each set. Each question shows its answer key and why that student gets it. Edit any question, add one or take one out, or press Write again with a note on what to do better. A set can change until its student starts.
- Publish, then add it to Canvas. Add to Canvas posts one assignment, a link to the quiz, graded complete or incomplete. It starts unpublished, so you choose when students see it.
- Students answer their own questions. Each one opens the assignment in Canvas and signs in with the email address Canvas has for them. Each answer is graded as they go, with feedback, score and the correct answer. One attempt, as practice.
- Mark completions. Mark completions in Canvas marks each student who turned it in as complete, after you preview it.
Nobody gets an email, and nothing runs on its own: the AI writes only when you start a quiz or press Write all again.
4. What the AI sees
- Each student's misses: the question, the right answer, and the choice or number they gave.
- Never their name, email or a sentence they wrote. A name reaches it only if you type one in a Write again note.
- The course's title and topics, so a student who missed little still gets a full set, on the topics nearest their misses.
Your department or teaching center may ask more. Personal quizzes for departments answers the usual questions on setup, your control and student data.
5. What it costs
Writing one student's set takes 50 AI credits, and a set that fails costs nothing. A class of 30 takes 1,500 credits, about $3. Answers are graded like any quiz's: choice questions and blanks graded from the answer key use no credits. Every new account gets 5,000 free credits, enough for 100 students' sets.
Tips for your first one
- Start it after a quiz closes and before the next unit. It aims at each student's worst and newest misses first.
- Read a few sets before you publish. If they miss the mark, press Write all again with one note for the whole class.
- Quiz questions that test one idea each give clearer misses, and so clearer follow-ups.