3dLME w/in subject voxel-wise covariate

AFNI Message Board
1

Hi All,

I am trying to model a single voxel-wise covariate the depends on the within subject factor. Is it that there is only one, one-level vVar allowed?

This is the error I received:

** Error:
Error with voxel-wise covariate BChange: Each subject is only
allowed to have one volume; that is, the covariate has to be at the
subject level.

My Script:

3dLME -prefix 3dLME_Valence+ALL
-mask “${rd}/scripts/group_mask_inter+tlrc.BRIK”
-model ‘Valence+Accuracy+RT+Age+CB+State+PosRate+NegRate+Bchange’
-vVars “BChange”
-qVars “Accuracy,RT,Age,CB,State,PosRate,NegRate”
-qVarCenters ‘0.87,829.81,21.73,1.50,34.14,6.59,2.50’
-ranEff ‘~1’
-SS_type 3
-jobs 6
-dataTable
Subj Valence Accuracy RT Age CB State PosRate NegRate BChange InputFile
3065 Positive 0.96 667.13 20.63 2 25 9 1 ${rd}/data/brain/orig/3065/3065_glm_bucket+tlrc.BRIK[78] ${rd}/data/brain/orig/3065/3065_glm_bucket+tlrc.BRIK[70]
…
3065 Negative 0.96 667.13 20.63 2 25 9 1 ${rd}/data/brain/orig/3065/3065_glm_bucket+tlrc.BRIK[74] ${rd}/data/brain/orig/3065/3065_glm_bucket+tlrc.BRIK[66]
…
3308 Negative 0.86 841.6 19.33 2 31 8 2 ${rd}/data/brain/orig/3308/3308_glm_bucket+tlrc.BRIK[74] ${rd}/data/brain/orig/3308/3308_glm_bucket+tlrc.BRIK[66]

2

Is it that there is only one, one-level vVar allowed?

Currently 3dLME only allows for one voxel-wise covariate per subject. The restriction can be loosened, but unfortunately I’m a little swamped right now.

3

I understand. It would eventually be helpful to have it added. I can create a difference score for now and include it across conditions. That should work the same, right?

4

I can create a difference score for now and include it across conditions. That should work the same, right?

Since your current model does not assume any interaction effect between Valence and BChange, so, yes, that should work fine under that assumption.

5

I’ve tried to include the same vVar file across both levels of Valence now so that there are not two levels. However, now it seems that the dimensions of the data frame of off?

I receive this error:

Error in dim(inData) ← c(dimx, dimy, dimz, lop$nVVars + lop$nSubj) :
dims [product 902629] do not match the length of object [59573514]
Execution halted

6

I’ve tried to include the same vVar file across both levels of Valence now so that there are not two levels.

If you want to use the difference score as a covariate, then you would also need to do the same for the conditions: use the contrast between the two conditions as input. That is, variable “Valence” would disappear from the model. And 3dttest++ would be good enough.