I collected pre-post data from a randomized controlled experiment. The intervention for the experiment group is an educational game and the pre-test and post-test contains 10 multiple-choice questions. The data is coded as 1= correct and 0=incorrect.
What is the right statistical test to analyse pre-post test?
Article ANOVA and ANCOVA of pre- and post-test, ordinal data
t-test or Wilcoxon Signed Rank
If you are not interested in the 10 different questions per se, you could calculate a sum score for the 10 questions, which would result in a sum-score ranging from 0 to 10. Do this appart for the pre- and post-tests.Then run an ANCOVA, i.e. a regression = post-sum-score = pre-sum-score + intervention/group
Probably you're interested in running a McNemar test. For dichotomous response options, I think it could fit your purpose. SPSS can do it. You'll find it in the crosstabs section. In statistics, select McNemar (and Risk Estimate to get some info about the odds ratio and the magnitude). This way is easier.
You can also do this analysis in the non-parametric tests tab, and then select 2 related samples.
Read this short paper to help you describe it properly in your method section:
Article Analysis of paired dichotomous data: A gentle introduction t...
Hope this was helpful!
Thank you. What about someone suggesting Linear Logistic Test Model or Linear Logistic Model for Change Pau Garcia Grau
Were the ten questions used at each stage parallel forms so that get x right on one means the same as x on the other or was something like choosing a random 10 for each person at stage one used. I am asking because if it was just two tests, or if the exact same test was used, there'd be some difficulties. There are a few choices, but some more info is necessary.
I used the exact 10 questions for the pre- and post- tests, but I changed the order of questions Daniel Wright
- Badges
- Science topic
- Similar topics
- Correction