Hi all,
I am performing a questionnaire analysis in SPSS using the Chi-square test (ordinal-ordinal, nominal-nominal, and ordinal-nominal comparison). However, my data does not meet the assumption of having at least 80% of the cells with expected count over 5.
I see that in this scenario the Likelihood ratio can be taken instead of the Pearson value. The issue is I am not able to find anywhere how to interpret it or report it, or even what it means exactly. Any help on that would be highly appreciated. Also please keep in mind that I am quite new at using these tests and analysing questionnaire data.
@John-Kåre ,
I don't use SPSS, but the following page suggests that you would need the Exact Tests module.
http://www-01.ibm.com/support/docview.wss?uid=swg21479647
In R, the standard fisher.test function handles r x c tables.
I suspect the problem could also be approached with Monte Carlo simulation or permutation tests.
The "asymp sig." is the two-sided p-value. This is for a Likelihood ratio test in the nominal-nominal case. It is interpreted just like a chi-square test of association. It is sometimes called a G-test.
I'm not sure that I agree that the likelihood ratio test is okay with low cell counts, though.
Fisher exact test is an option in this case, although it does have an assumption of fixed margins.
You might also determine if SPSS automatically --- or as an option --- applies Yates or Williams continuity correction, which help in cases of low cell counts.
A chi-square might not be appropriate here. I suggest you use Fisher's exact test.
Chi square test should not be used if expected frequency of any cell is less than 5.
Fisher test can only be used with 2x2 tables. Based on your description, it seems like you have variables with many categories (your output says you have 30(!) cells with expected count
@John-Kåre , some software packages can conduct an exact test on tables larger than 2 x 2. This includes SPSS (that the OP is using) and R.
That's beyond me, @Salvatore. You don't get Fisher Exact test values in SPSS output if you have more than 2 x 2 table. How can you obtain them?
jk
@John-Kåre ,
I don't use SPSS, but the following page suggests that you would need the Exact Tests module.
http://www-01.ibm.com/support/docview.wss?uid=swg21479647
In R, the standard fisher.test function handles r x c tables.
I suspect the problem could also be approached with Monte Carlo simulation or permutation tests.
Thanks, @Salvatore.
Nice to know! Sadly, I'm not familiar with R.
Best
JK
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics