3dMVM or 3dLME?

AFNI Message Board
1

I am here to ask for advice on whether 3dMVM or 3dLME is the better option for our data analysis. The data consist of 2 between subjects factors (Group category: Group 1 vs Group 2; Ratings: continuous between 0 - 100, one value per participant) and 1 within subjects factor (9 Time points estimated from the TENT function). I would like to perform an analysis to test whether the TENT shape is differs based on the participants’ Group and Rating (Time x Group x Rating interaction). As an initial approach I’ve set up this analysis in 3dMVM (code copied below), since I have 1 within subjects factor and the between subjects factors do not differ across levels of the within subjects factor. Is this the advisable route to take or is 3dLME better for this purpose?


#!/bin/sh

3dMVM -prefix ANCOVA_TENT0to9 -jobs 6 \
-bsVars group*rate \
-qVars rate \
-wsVars time \
-dataTable @ANCOVAtable.txt
2

3dMVM is perfectly fine in your case.

3

thank you!

I also want to test the Group x Rating effect separately at each time point. For this purpose I am using the glfCode option - for example at the first time point I use the symbolic coding:


-num_glf 2 -glfLabel 2 vbias_group_time0 -glfCode 2 group : '1*Group1' '-1*Group2' time : '1*Time_0' vbias : \

Is this method the advised way of testing the between subjects interactions (which includes on continuous variable) at each level of a single within subjects variable?

4

-num_glf 2 -glfLabel 2 vbias_group_time0 -glfCode 2 group : ‘1Group1’ '-1Group2’ time : ‘1*Time_0’ vbias : \

Use glt instead (and remove some single quotes in the coding):

-num_glt 2 -gltLabel 2 vbias_group_time0 -gltCode 2 group : ‘1Group1 -1Group2’ time : ‘1*Time_0’ vbias : \