Which statistical test is appropriate if we want to look for statistical differences between total counts for each category, if the subjects can select multiple categories in their response?
The survey question asks: "Did you make changes to any of the following?"
a) Instruction
b) Student experience
c) Teacher responsibilities
Subjects can select more than one category (e.g. they can select Instruction and Student experience). Therefore, for each subject, we have a 0 or 1 for each category in our data. We have summed the counts for each category across all subjects and want to evaluate, for example, whether significantly more subjects mentioned Instruction than Student experience.
Hi Helena,
you might use a chi² test, whose null hypothesis claims that there are no differences between the frequencies of each category. You can do it either pairwise (three tests comparing two categories) or (maybe even better) by comparing all three groups simultaneously.
Best,
Brian
Hello Helena,
The three cells (or six, if you envision coding each person's response as "yes" or "no" for each of the three options) are not mutually exclusive, as each respondent can say "yes" (or "no") to all three. So, on reflection, the chi-square test is likely not the best choice.
For pairs of options, the McNemar test would allow you to compare whether significant differences exist.
Good luck with your work.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey