I'm running a logistic mixed-effects model in which I predict word knowledge from condition (three levels: experimental1, experimental2, control), time of testing (immediately, delayed), and the interaction between condition and testing moment.
The model output looks like this:
Fixed effects Estimate (logit) Std. error z-value p-value (Intercept) -2.32 0.40 -5.78 < .001 Condition:Experimental1 0.96 0.30 3.16 .002 Condition:Experimental2 0.78 0.31 2.52 .01 Testing moment:Delayed -0.04 0.10 -0.46 .65 Cond:Exp1 * Test:Delay -0.09 0.13 -0.72 .47 Cond:Exp2 * Test:Delay -0.16 0.14 -1.22 .22 Random effects Variance Std. dev. Participant (intercept) 0.93 0.97 Word (intercept) 1.85 1.36
I can see that the contrast Experimental1 - Control has a p-value of .002, and the contrast Experimental2 - Control has a p-value of .01. Through relevelling, I found that the p-value for the contrast Experimental1 - Experimental 2 is .56.
My question is whether any correction for multiple testing should be applied. For the factor of condition, should alpha be .05 / 3 = .0167?
If so, does alpha stay at .05 for Testing moment, where only one comparison can be made? (immediate - delayed)