post-hoc analysis of two-way repeated measures ANOVA and multcompare() row estimate of Friedman stats

Is there a different way of performing a post-hoc analysis on a two-way repeated measures ANOVA other than running a whole bunch of paired t-tests or signrank() tests? Because i've got 2 variables: condition (visual, auditory and audiovisual) and difficulty (easy, intermediate and hard) so a total of 9 different combinations. If i compare all of them to eachother the bonferonni corrected significance level would be 0.05/36 = 0.0014 and that's just too low.

Also, im trying to find out which means per difficulty are significantly different from eachother by using multcompare() on the Friedman stats, but the documentation of multcompare() says that only the columns can be compared. However, the difficulty levels are represented by the rows.

With regards to the multcompare() function, the name-value pair 'Estimate', 'row' is available for the stats resulting from anova2(), but not for the non-parametric two-way repeated measures friedman() analysis. Why?

Answer 1

[neuromechanist]

Hi,

I believe for repeated mesure ANOVA, you might need to use ranova() and not anova2(). However, this function works best if you use tables. You are right, the design of ranova (and multcompare) is based on columns, but once you set up your table, everything goes very smmoothly.