3dLME Group by Time Effects

AFNI Message Board
1

Dear AFNI experts,

Hope all of you are staying safe. I am running 3dLME and am wondering how can I analyze Group by Time effects. I have two groups of subjects (A and B) and both groups have before (pre) and after (post) data. Below is the script I used.

3dLME -prefix 3dLME -jobs 5
-model ‘GroupTime’ -ranEff ‘~1’ -SS_type 3
-num_glt 5
-gltLabel 1 ‘Apost-Apre’ -gltCode 1 'Group : 1A Time : 1post -1pre’
-gltLabel 2 ‘Bpost-Bpre’ -gltCode 2 ‘Group : 1B Time : 1post -1pre’
-gltLabel 3 ‘Apre-Bpre’ -gltCode 3 'Group : 1A -1B Time : 1pre’
-gltLabel 4 ‘Apost-Bpost’ -gltCode 4 ‘Group : 1A -1B Time : 1post’
-gltLabel 5 ‘Post-Pre’ -gltCode 5 'Time : 1post -1*pre’
-mask GMmask+tlrc
-dataTable
Subj Group Time InputFile
0001 A pre 0001.pre/clusterave1_r2z+tlrc
0001 A post 0001.post/clusterave1_r2z+tlrc
0002 B pre 0002.pre/clusterave1_r2z+tlrc
0002 B post 0002.post/clusterave1_r2z+tlrc
.
.
.
.
.

I could see Group:Time F results in the output data. Could you please let me know if this is the Group by Time effect I am looking for?

Best,
JW

1.jpg
1.jpg1152×648 131 KB

2

Could you please let me know if this is the Group by Time effect I am looking for?

Yes, that’s the conventional omnibus interaction effect between Group and Time in your case. It can be better understood when combined with the four partitions you specified:

-gltLabel 1 ‘Apost-Apre’ -gltCode 1 ‘Group : 1A Time : 1post -1pre’
-gltLabel 2 ‘Bpost-Bpre’ -gltCode 2 'Group : 1B Time : 1post -1pre’
-gltLabel 3 ‘Apre-Bpre’ -gltCode 3 ‘Group : 1A -1B Time : 1pre’
-gltLabel 4 ‘Apost-Bpost’ -gltCode 4 'Group : 1A -1B Time : 1post’ \

3

Hi Gang,

Thank you! Could you please let me know if you meant to remove this line?
-gltLabel 5 ‘Post-Pre’ -gltCode 5 ‘Time : 1post -1pre’ \

Also, could you please let me know if it is correct to have Group : Time F on both OLay and Thr?

Best,
JW

4

Could you please let me know if you meant to remove this line?
-gltLabel 5 ‘Post-Pre’ -gltCode 5 ‘Time : 1post -1pre’ \

No, not at all. I was only trying to point out the other four comparisons are directly associated with the omnibus F-statistic for the interaction. The 5th contrast serves its own purpose if it’s what you intended to know.

Also, could you please let me know if it is correct to have Group : Time F on both OLay and Thr?

Sure, that’s fine because there is no single effect estimate associated with an F-statistic in general. However, since you have a special case of 2 x 2 design, you could add another post hoc test:

‘Group : 1A -1B Time : 1post -1pre’

Then you can set the effect estimate as OLay.

5

Hi Gang,

Thank you for the instruction. I’ll try that. Stay safe!

Best,
JW