Dear AFNI Gurus,

I am trying to set up 3dANOVA3 –type 5 to do a split plot mixed effects analysis with the following parameters:

We have 14 controls and 14 patients, but for the sake of brevity we only included 3 subjects per group in the syntax below. The design is to 2 runs of the same task separated by a manipulation. The hypotheses are (1) no difference between groups at run1, (2) significant differences between groups at run2. Thus we are predicting a group by run interaction. What would be the best way of setting up the syntax below to examine these hypotheses?

#a=group (1=ctrl, 2=patient); fixed
#b=run (1=run1, 2=run2); fixed
#c=subjects (3 per group); random

3dANOVA3 \
-type 5 \
-alevels 2 -blevels 2 -clevels 3 \

-dset 1 1 1 ../../ctrl1/analysis/NegvPosRun1+tlrc'[0]' \
-dset 1 1 2 ../../ctrl2/analysis/NegvPosRun1+tlrc'[0]' \
-dset 1 1 3 ../../ctrl3/analysis/NegvPosRun1+tlrc'[0]' \

-dset 1 2 1 ../../104/analysis/NegvPosRun2+tlrc'[0]' \
-dset 1 2 2 ../../105/analysis/NegvPosRun2+tlrc'[0]' \
-dset 1 2 3 ../../106/analysis/NegvPosRun2+tlrc'[0]' \

-dset 2 1 1 ../../101/analysis/NegvPos101s1PANN+tlrc'[0]' \
-dset 2 1 2 ../../102/analysis/NegvPos102s1PANN+tlrc'[0]' \
-dset 2 1 3 ../../103/analysis/NegvPos103s1PANN+tlrc'[0]' \

-dset 2 2 1 ../../101/analysis/NegvPos101s2PANN+tlrc'[0]' \
-dset 2 2 2 ../../102/analysis/NegvPos102s2PANN+tlrc'[0]' \
-dset 2 2 3 ../../103/analysis/NegvPos103s2PANN+tlrc'[0]' \

-diskspace \
-fa GrpEff \
-fb RunEff \
-fab GrpxRun \
-amean 1 ctrl \
-amean 2 exp \
-bmean 1 run1 \
-bmean 2 run2 \
-bucket ANOVA_SRET_NegvPos

-fab GrpxRun should give us the interaction. In addition, how do we use the “diff” and “contrast” options to get the between group comparisons at run1 and run2?

Alternatively, should we code factor b with 4 levels as follows:

1 - Factor 1 (level 1); Factor 2 (level 1) control run1
2 - Factor 1 (level 1); Factor 2 (level 2) control run2
3 - Factor 1 (level 2); Factor 2 (level 1) patients run1
4 - Factor 1 (level 2); Factor 2 (level 2) patients run2

3dANOVA3 \
-type 5 \
-alevels 2 -blevels 4 -clevels 3 \

-dset 1 1 1 ../../ctrl1/analysis/NegvPosRun1+tlrc'[0]' \
-dset 1 1 2 ../../ctrl2/analysis/NegvPosRun1+tlrc'[0]' \
-dset 1 1 3 ../../ctrl3/analysis/NegvPos Run1+tlrc'[0]' \

-dset 1 2 1 ../../ctrl1/analysis/NegvPosRun2+tlrc'[0]' \
-dset 1 2 2 ../../ctrl2/analysis/NegvPosRun2+tlrc'[0]' \
-dset 1 2 3 ../../ctrl3/analysis/NegvPosRun2+tlrc'[0]' \

-dset 2 3 1 ../../patient1/analysis/NegvPosRun1+tlrc'[0]' \
-dset 2 3 2 ../../patient2/analysis/NegvPosRun1+tlrc'[0]' \
-dset 2 3 3 ../../patient3/analysis/NegvPosRun1+tlrc'[0]' \

-dset 2 4 1 ../../patient1/analysis/NegvPosRun2+tlrc'[0]' \
-dset 2 4 2 ../../patient2/analysis/NegvPosRun2+tlrc'[0]' \
-dset 2 4 3 ../../patient3/analysis/NegvPosRun2+tlrc'[0]' \

-diskspace \
-fa GrpEff \
-fb RunEff \
-fab GrpxRun \
-amean 1 ctrl \
-amean 2 patient \
-bmean 1 control run1 \
-bmean 2 control run2 \
-bmean 3 patients run1 \
-bmean 4 patients run2 \
-bucket ANOVA_SRET_NegvPos

How do I set up “diff” or “contrast” to get
1. run1 controls vs patients
2. run 2 controls vs patients
3. the interaction

One idea is to do the following:

run1 controls vs patients:
1 0 -1 0

run 2 controls vs patients:
0 1 0 -1

Interaction Factor 1 by Factor2:
1 -1 -1 1

Also, this entire analysis is for the Neg vs Pos contrast. Is there a way to add another factor that would examine the BOLD response for multiple contrasts, namely, Neg vs Neutral, Pos vs. Neutral, and Neg vs. Pos contrasts. Or do we have to run 3dANOVA3 three separate times – once for each contrast?

Many thanks,
Philippe and Wiveka, AFNI neophytes/wizards in training!