Interpreting output of behavioural covariate

AFNI Message Board
1

I am attempting to look at how a behavioural covariate is related to differences in brain activation.
I have a number of glt terms that look at how my quantitative covariate relates to various various conditions and/or contrasts.

But in the output, the maps that include the covariate are, in many cases, identical (both the coefficient maps and the t maps)
My design has two time points, and 2x2 conditions of lexicality (lex) and first/second languages (L1L2).

So the two following glt terms are meant to be looking at how my covariate (a individual measure of sensitivity to word frequency) relates to activation for words in each language


-gltCode L1:wrd:freq 'L1L2 : 1*L1 Lex : 1*wrd freq : ' \
-gltCode L2:wrd:freq 'L1L2 : 1*L2 Lex : 1*wrd freq : ' \

Interestingly, other glts also return identical maps, but ones that are different from the above, so for example the following all return identical maps:


-gltCode L1:prnt:psd:freq 'L1L2 : 1*L1 Lex : 1*psd freq : ' \
-gltCode L2:prnt:psd:freq 'L1L2 : 1*L2 Lex : 1*psd freq : ' \
-gltCode L1:psd:t1:freq 'L1L2 : 1*L1 Lex : 1*psd Time : 1*Time1 freq : ' \
-gltCode L2:psd:t1:freq 'L1L2 : 1*L2 Lex : 1*psd Time : 1*Time1 freq : ' \
-gltCode L1:psd:t3:freq 'L1L2 : 1*L1 Lex : 1*psd Time : 1*Time3 freq : ' \
-gltCode L2:psd:t3:freq 'L1L2 : 1*L2 Lex : 1*psd Time : 1*Time3 freq : ' \

Am I misunderstanding the interpretation of the covariate and what it means in glts such as these?
Thanks

2

Could you show your script?

3

Attached below. As described, ‘Time’, ‘Lex’, and ‘L1L2’ are the 2x2x2 within-subject conditions. ‘freq’ is the continuous behavioural covariate.
The activation maps without the covariate (i.e. the glt following each glt with the covariate) are not identical, i.e the output for


-gltCode L1:wrd 'L1L2 : 1*L1  Lex : 1*wrd ' \

is not identical to the output for


-gltCode L2:wrd 'L1L2 : 1*L2  Lex : 1*wrd ' \

but the output for


-gltCode L1:wrd:freq 'L1L2 : 1*L1  Lex : 1*wrd freq : ' \

is identical to that for


-gltCode L2:wrd:freq 'L1L2 : 1*L2  Lex : 1*wrd freq : ' \

So I know that this is not simple a problem with the datatable mistakenly having the same files for both L1 and L2, for example.
The script is as follows:



#!/bin/bash
																
3dLMEr -prefix output_lmer_freq \
-jobs 4 \
-mask TT_Mask.nii.gz \
-model 'Time*Lex*L1L2+freq+(1+Time+L1L2|Subj)' \
-qVars 'freq' \
\
-gltCode L1:wrd:freq 'L1L2 : 1*L1  Lex : 1*wrd freq : ' \
-gltCode L1:wrd 'L1L2 : 1*L1  Lex : 1*wrd ' \
-gltCode L2:wrd:freq 'L1L2 : 1*L2  Lex : 1*wrd freq : ' \
-gltCode L2:wrd 'L1L2 : 1*L2  Lex : 1*wrd ' \
\
-gltCode L1:wrd:t1:freq 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time1 freq : ' \
-gltCode L1:wrd:t1 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time1 ' \
-gltCode L2:wrd:t1:freq 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time1 freq : ' \
-gltCode L2:wrd:t1 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time1 ' \
\
-gltCode L1:wrd:t3:freq 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time3 freq : ' \
-gltCode L1:wrd:t3 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time3 ' \
-gltCode L2:wrd:t3:freq 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time3 freq : ' \
-gltCode L2:wrd:t3 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time3 ' \
\
-gltCode L1:wrd:t3-t1:freq 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L1:wrd:t3-t1 'L1L2 : 1*L1  Lex : 1*wrd Time : 1*Time3 -1*Time1 ' \
-gltCode L2:wrd:t3-t1:freq 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L2:wrd:t3-t1 'L1L2 : 1*L2  Lex : 1*wrd Time : 1*Time3 -1*Time1 ' \
\
-gltCode L1:psd:freq 'L1L2 : 1*L1  Lex : 1*psd freq : ' \
-gltCode L1:psd 'L1L2 : 1*L1  Lex : 1*psd ' \
-gltCode L2:psd:freq 'L1L2 : 1*L2 Mod : 1*prnt Lex : 1*psd freq :	' \
-gltCode L2:psd 'L1L2 : 1*L2 Mod : 1*prnt Lex : 1*psd ' \
\
-gltCode L1:psd:t1:freq 'L1L2 : 1*L1 Mod : 1*prnt Lex : 1*psd Time : 1*Time1 freq : ' \
-gltCode L1:psd:t1 'L1L2 : 1*L1  Lex : 1*psd Time : 1*Time1 ' \
-gltCode L2:psd:t1:freq 'L1L2 : 1*L2  Lex : 1*psd Time : 1*Time1 freq : ' \
-gltCode L2:psd:t1	'L1L2 : 1*L2  Lex : 1*psd Time : 1*Time1 ' \
\
-gltCode L1:psd:t3:freq 'L1L2 : 1*L1  Lex : 1*psd Time : 1*Time3 freq : ' \
-gltCode L1:psd:t3 'L1L2 : 1*L1  Lex : 1*psd Time : 1*Time3 ' \
-gltCode L2:psd:t3:freq 'L1L2 : 1*L2  Lex : 1*psd Time	: 1*Time3 freq : ' \
-gltCode L2:psd:t3 'L1L2 : 1*L2  Lex : 1*psd Time : 1*Time3 ' \
\
-gltCode L1:psd:t3-t1:freq 'L1L2 : 1*L1  Lex : 1*psd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L1:psd:t3-t1 'L1L2 : 1*L1  Lex : 1*psd Time : 1*Time3 -1*Time1 ' \
-gltCode L2:psd:t3-t1:freq 'L1L2 : 1*L2  Lex : 1*psd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L2:psd:t3-t1 'L1L2 : 1*L2  Lex : 1*psd Time : 1*Time3 -1*Time1 ' \
\
-gltCode L1:psd-wrd:freq 'L1L2 : 1*L1  Lex : 1*psd -1*wrd freq : ' \
-gltCode L1:psd-wrd 'L1L2 : 1*L1  Lex : 1*psd -1*wrd ' \
-gltCode L2:psd-wrd:freq 'L1L2 : 1*L2 Mod : 1*prnt Lex : 1*psd -1*wrd freq :	' \
-gltCode L2:psd-wrd 'L1L2 : 1*L2 Mod : 1*prnt Lex : 1*psd -1*wrd ' \
\
-gltCode L1:psd-wrd:t1:freq 'L1L2 : 1*L1 Mod : 1*prnt Lex : 1*psd -1*wrd Time : 1*Time1 freq : ' \
-gltCode L1:psd-wrd:t1 'L1L2 : 1*L1  Lex : 1*psd -1*wrd Time : 1*Time1 ' \
-gltCode L2:psd-wrd:t1:freq 'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time1 freq : ' \
-gltCode L2:psd-wrd:t1	'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time1 ' \
\
-gltCode L1:psd-wrd:t3:freq 'L1L2 : 1*L1  Lex : 1*psd -1*wrd Time : 1*Time3 freq : ' \
-gltCode L1:psd-wrd:t3 'L1L2 : 1*L1  Lex : 1*psd -1*wrd Time : 1*Time3 ' \
-gltCode L2:psd-wrd:t3:freq 'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time3 freq : ' \
-gltCode L2:psd-wrd:t3 'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time3 ' \
\
-gltCode L1:psd-wrd:t3-t1:freq 'L1L2 : 1*L1  Lex : 1*psd -1*wrd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L1:psd-wrd:t3-t1 'L1L2 : 1*L1  Lex : 1*psd -1*wrd Time : 1*Time3 -1*Time1 ' \
-gltCode L2:psd-wrd:t3-t1:freq 'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time3 -1*Time1 freq : ' \
-gltCode L2:psd-wrd:t3-t1 'L1L2 : 1*L2  Lex : 1*psd -1*wrd Time : 1*Time3 -1*Time1 ' \
-dataTable @/Volumes/a187-rsync/group/2019_MVM/bb1/datatable.txt
4

In your model specification,

-model ‘TimeLexL1L2+freq+(1+Time+L1L2|Subj)’ \

you assumed that the effect of ‘freq’ is independent of ‘Time’, ‘Lex’ and ‘L1L2’; therefor, it is no surprise that you see those identical results. If you expect to have different ‘freq’ effect between ‘L1’ and ‘L2’, change your model to

-model 'TimeLexL1L2+L1L2*freq+…"

Your random-effects specification “(1+Time+L1L2|Subj)” is a little strange. It’s probably better to have “(1+Time|Subj)+(1+L1L2|Subj)”