Significance testing proportions when there are more than two subgroups

#1
I have asked a series of questions on a 5 point rating scale (strongly disagree to strongly agree) and would like to test if there are any significant differences between my 5 subgroups. I know that I can test the mean of each using ANOVA and this will tell me if all means are equal or not, and if any of the subgroups are statistically significant to any others it will detail which. However what's the appropriate test if I want to test e.g. the % saying strongly agree between each subgroup, in order to get the detail on which subgroup is sig different to which other subgroup(s)?
Any help would be greatly appreciated, thank you.
 

Karabiner

TS Contributor
#2
Some would argue that a 5-point rating scale is ordinal and that you should use Kruskal-Wallis H test, not analysis of variance.

But anyway, regarding your question, you transform your rating scale into a dichotomous variable ("strongly agree vs. all other levels"),
so you have a 5*2 table and can use the Chi² test.

With kind regards

Karabiner
 
Last edited:
#3
Some would argie that a 5-point rating scale is ordinal and that you should use Kruskal-Wallis H test, not analysis of variance.

But anyway, regarding your question, you transform your rating scale in a dichotomous variable ("strongly agree vs. all other levels"),
so you have a 5*2 table and can use the Chi² test.

With kind regards

Karabiner
Thank you so much for your help. Assuming that my subgroups are large enough, could I run on a 5*5 table - so I can pull out significant differences on strongly agree and then separately on 'agree' etc? Which post hoc test would identify where the differences are? Thanks again