My data is nominal/categorical. I want to see if 4 groups differ significantly. The question for them is if they convict a person or if they acquit the person. So the data is categorical: convict or acquit.
With the chi-square test I can only see that the difference between all of the four groups is significance, but I want to see for example if the difference between group 1 and 2 is also significance and between group 2 and 3 etc. I know this is possible with the Kruskal Wallis test, only I read online that this test is only for ordinal data.
I hope someone can help me out. I am sorry, my English is not really good.
Julia Kersbergen Of course. Select "Analyze->Descriptive Statistics->Crosstabs...". Insert your two variables as row and column respectively. Under 'Statistics', select 'Chi-square'. Under 'Cells', select 'Compare column proportions' and 'Adjust p-values (Bonferroni method)' (you can also select which counts and percentages to display in your table, feel free to try out these options in order to obtain what you need to display). That's it, press Ok. A crosstabulation table will be displayed, and every cell will be marked with a letter, 'a', 'b', 'c', etc... Cells containing the same letter are not different in proportions, the others contains different proportions.
Hi,
one way to determine whether each group differs from the others after a Chi-Square test is to run post hoc multiple z-tests of two proportions. This way you will have pairwise comparisons among all groups. The procedure is automated in many statistical softwares (like IBM SPSS), and p-values can be adjusted using Bonferroni method.
Emanuele Cassioli thank you for your answer. Can you tell me how I can run post hoc z-tests in SPSS? And also the Bonferroni method? I don't use SPSS much so I don't know much about it. Thank you!!
Julia Kersbergen Of course. Select "Analyze->Descriptive Statistics->Crosstabs...". Insert your two variables as row and column respectively. Under 'Statistics', select 'Chi-square'. Under 'Cells', select 'Compare column proportions' and 'Adjust p-values (Bonferroni method)' (you can also select which counts and percentages to display in your table, feel free to try out these options in order to obtain what you need to display). That's it, press Ok. A crosstabulation table will be displayed, and every cell will be marked with a letter, 'a', 'b', 'c', etc... Cells containing the same letter are not different in proportions, the others contains different proportions.
Emanuele Cassioli Thank you for your useful recommendation. Can we use the same method if we are running a Fisher exact test when our sample is not large enough for the Chi-square test?
Farzad Radmehr Yes you can. Multiple z-tests are simply a way to compare every possible pair of groups, they can be performed primarily or after any other test of choice.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running