# Post hoc test for robust 3 way ANOVA

I am doing a 3x2x2 between subject study. To make thing simple, I name my variables as A, B and C here. Using the WRS 2 package in R, I have gotten results for the 3 way robust ANOVA (t3way). One of the two-way interactions and the three way interaction were significant. The t3way anova result showed a significant result between AB and ABC with p-value of .031 and .051 respectively.

When come to the post hoc test, I was made understood that there is no post hoc test available for 3 way robust ANOVA in R and hence, I have to split the file by the third IV for the two way ANOVA and post hoc tests. I split the data with variable C (i.e. Ci and Cii) and the t2way anova result of Ci is:

A .394(p-value)

B .618

AB .019

t2way anova for Cii din't show any significant.

When I performed mcp2atm (post hoc test) on the separate robust 2 way ANOVAs, I didn't get a single contrast that is significant. Post hoc test (mcp2atm) result for Ci showed:

    V1       ci.lower ci.upper p-value


A1 1.00000 -1.11291 3.11291 0.22754

A2 1.33333 -0.87160 3.53827 0.12861

A3 0.33333 -2.56160 3.22827 0.77417

B1 -0.33333 -2.68024 2.01358 0.77417

A1:B1 0.33333 -1.77957 2.44624 0.68124

A2:B1 0.00000 -2.20493 2.20493 1.00000

A3:B1 -0.33333 -3.22827 2.56160 0.77417

I am puzzled and do not know how to continue with the analysis.

• Can you paste in your output? A significant omnibus test without a significant post-hoc isn't necessarily unusual, but it's going to be hard to explain without more information. – gung Jul 7 '15 at 10:56
• Please edit your Q & paste the output into it. – gung Jul 7 '15 at 12:06