I implemented a pretest-posttest on testing student spatial ability. I have the items categorized into specific aspects and would like to compare the number of students between tests who got each item correct. For example, for question #1, I have an 87% correct response rate for the pretest and 82% for the posttest.
My sample size is 180 and I have been thinking of using a paired t-test for each item, but would that be okay considering I'm not necessarily comparing "means"?
Obviously, t-tests are not a good answer. Because you do not have a mean or standard deviation.
You can easily use the Two-Proportion Test in Minitab software.
Abolfazl Ghoodjani Thank you for this. I would like to follow if the two-proportion z-test would still be valid even if my sample sizes for the two groups are the same? All same 180 students answered the pretest and posttest. I would just like to clarify since all examples I have seen involve two different population sizes. :)
Yes, it is still valid and there is no need for the size of samples to be equal in the two groups.
Jose Manansala , the first question is if you have data recorded so that you can match an individual student's pre- result to their post- result. And if so, if you think the relevant way look at the problem is that you want to compare how many students got the question wrong in the pre- and got the question right in the post-, and compare this to how many students got the question right in the pre- and got the question wrong in the post-. (If that description makes sense).
If this is what you want, then McNemar's test works.
If you didn't record each students result pre- and post- in a way you can identify the student, or don't want to consider this, then a two-sample z test of proportion is appropriate.
This test is similar or equivalent to a chi-square test of association, depending on the options chosen. And the chi-square may be easier or more avialable depending on what software you use.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size