Intro to Statistics at IU Indy - Two Samples t-Tests: Paired-Samples
Overview
Andy Baker explains paired-samples t-tests as tests of whether the population mean difference between matched measurements is zero, with each difference calculated consistently as after minus before. He works through the difference-score standard deviation, standard error, t statistic, degrees of freedom, critical-value decision, and confidence interval, then applies the method to stress ratings and study hours.
Key takeaways
- A paired-samples t-test operates on one difference score per matched pair, calculated in a consistent direction such as after minus before.
- The paired-test null hypothesis is that the population mean difference is zero; the sign of the sample mean difference indicates the direction of observed change.
- For n paired observations, the standard error is the sample standard deviation of the differences divided by √n, and the test uses n − 1 degrees of freedom.
- With six pairs and df = 5, a two-tailed test at the 5% level uses critical values of approximately ±2.571.
- The stress example gives t ≈ −1.17 and an approximate interval from −1.36 to 0.36, so it does not provide clear evidence of a mean change.
- The midterm study example gives t ≈ 4.47 and an interval from about 0.843 to 3.157 hours, indicating greater average study time during midterm week.
Chapters
- Define each paired difference D as after minus before, keeping the order consistent; a change from 150 lb to 300 lb is D = 150 lb.
- The null hypothesis is that the population mean difference equals zero, while the two-sided alternative says it is not zero.
- For six example differences with mean 2 and sample standard deviation about 1.095, the standard error is 1.095 divided by √6, or about 0.447.
- A positive mean difference indicates an average increase, while a negative mean difference indicates an average decrease.
- Calculate the paired-samples statistic as mean difference divided by its standard error, with degrees of freedom equal to n − 1.
- For n = 6, df = 5 and the two-tailed 5% critical values are approximately −2.571 and +2.571.
- In the stress example, the mean change is −0.5, the standard deviation is 1.05, the standard error is about 0.429, and t is about −1.17.
- Because −1.17 falls between the critical values, the test fails to reject no average change; the approximate confidence interval from −1.36 to 0.36 also contains zero.
- Six students’ midterm-week study hours are compared with their typical-week hours, using midterm week minus typical week as the difference.
- With a mean difference of 2 hours and a difference-score standard deviation of 1.10, the standard error is about 0.45 hours.
- The resulting t statistic is about 4.47 with 5 degrees of freedom, exceeding the two-tailed critical value of 2.571.
- The test rejects a zero mean difference; the confidence interval is approximately 0.843 to 3.157 hours, supporting the conclusion that students studied more during midterm week.
Transcript excerpt
0:05 If you could choose any number out there, any number that would be the best to describe no change or no difference. I think the best one would be zero. That makes sense, right? And that's going to come up here in our null hypothesis. And so we're assuming no change, no difference between before and after, no difference between husband and wife, no difference between one twin and the other. And so our null hypothesis is going to be that the mean differences, the average difference is going to be equal to zero. No change, no difference. Whereas the alternative hypothesis is that the mean differences is not going to be zero. That there is going to be something different. And so
0:52 this is saying that the average difference in the population is zero. Any observed changes is just essentially anything observed is just noise. Our alternative hypothesis says that the average difference is not zero. And so we're going to assume no real change until our data says something different. The key for the paired samples is the difference score or essentially D. And so each of you, each person is given one number in terms of the result. And all that is is just the after minus the before. It's got to be that order. So the after minus and of course for the
1:40 husband, wife, the twins, this isn't going to matter as much. But the after, you have to keep it consistent whatever you do. the after minus the before. And so let's say I've been playing tennis. Uh and so because I'm playing tennis, I I now can bench press 300 lb. Before I could only bench 150 lb. And so after minus before and so that's then 150 would be my D equals my
This is the start of the transcript. Save this video free for the whole transcript, searchable and time-coded.
Related notes
Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, Andy Baker.