Could you direct me to a software (SAS, SPSS, ... etc.) in which it is possible to ask for post-hoc tests for chi-square tests for goodness of fit or for the analysis of contingency tables? Thanks.
If by post hoc you mean several specific comparison tests, each with their own p values, then one of the simplest approach is to calculate the individual p values and use the p.adjust function in R. Say your individual p values are .01,.1,.2.,3,.02,.005, you type
p.adjust(c(.01,.1,.2,.3,.02,.005))
and R produces
[1] 0.05 0.30 0.40 0.40 0.08 0.03
which are those adjusted using Holm's method, which is the default.
Alternatively, you might first use different methods for plotting the residuals (like correspondence analysis). Given that if there are a lot of categories in each variable there will be lots of possible comparisons (and therefore lots of p values), it is not clear to me how to interpret that many p values (with or without adjustment).
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Software