Currently I have a dataset with students from 8 different studies identifying which values (Responsibility, Trust, etc.) they find important. They were allowed to choose up to 5 values from 17. I defined a Multiple Response Set from the 17 Values via "Analyze/Tables/Multiple Response Sets" and after that performed a Chi² analysis via "Analyze/Tables/Custom Tables". Unfortunately, I have several (>20%) cells with an expected value lower than 5. Consequently, I am strugling finding a way to perform an exact test like the Fisher's exact test on my 8X17 continguency table.
I found several threads on exact tests for larger continguency tables or analyzing multiple response sets, but not the combination of the two. SPSS does provide this option via the 'standard way' of performing a Chi² test using the Exact Toolbox, but not for multiple response sets as far as I know. Can anyone help me out on this one? Thank you very much in advance!
Best,
Arjan
This tutorial may help you:
https://www.youtube.com/watch?v=aLp91GoPhDs
I hope you helpful
Dear Gioacchino,
Thank you for your help. However, the video is only about the Chi² test for multiple responses. The method shown doesn't provide me with the opportunity to perform exact tests. I was hoping to find a way to perform an exact test on an m X n continguency table from a multiple response set.
Dear Arjan,
in such case you should operate by using syntax of SPSS, but it comes to impute a code with many strings.
I do not know how else to help you through these posts.
Assuming you're speaking of the usual independence of homogeneity chi-square test — Use fisher.test in R. If it saturates the memory, use the simulate = TRUE option (and if you're going this way, you can as well use simulate = TRUE in chisq.test).
However, if a given individual can select more than one value among the 17, this test does not apply, and you do not have a 8×17 contingency table. You may have a 8×17×2 contingency table, for which a chi-square test may be built but an exact test would be more difficult. Simulations to get the (approximate) true p-value may be then a hint.
Hi Arjan,
If understand your question, have you tried desirability function analysis for multiple response in R package? this might help,
https://onlinecourses.science.psu.edu/stat503/node/78
http://www.itl.nist.gov/div898/handbook/pri/section5/pri5322.htm
Regards
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS