3dMVM error GLT

AFNI Message Board
1

AFNI version info (afni -ver):

Dear experts,

I'm using 3dMVM to calculate the interaction effect of two within variables. The first one is the Time variable (pre and post treatment), the second one is a three-level varaible which represents three parameter of a function (HRF function of BOLD signal).
I want to explore the effect of time separately on the three parameters of the HRF (time-to-peak, height, and fwhm).

This is the code for 3dMVM:

3dMVM -prefix Timing_Effect_on_HRFparam \
      -bsVars "Sex+Age" \
      -wsVars "Timing+HRFparam" \
      -qVars "Age" \
      -mask mean_all_mask.nii \
      -num_glt 4 \
      -gltLabel 1 'Pre_vs_Post' -gltCode 1 'Timing : 1*pre -1*post' \
      -gltLabel 2 'Pre_vs_Post_on_height' -gltCode 2 'Timing : 1*pre -1*post * HRFparam : 1*height' \
      -gltLabel 3 'Pre_vs_Post_on_timetopeak' -gltCode 3 'Timing : 1*pre -1*post * HRFparam : 1*time' \
      -gltLabel 4 'Pre_vs_Post_on_FWHM' -gltCode 4 'Timing : 1*pre -1*post * HRFparam : 1*fwhm' \
      -dataTable @dataTable.txt

I have already run analyses with the following code for 3dMVM, but this time this error came out:

"** Error:
   Failed at GLT No. 1! Make sure that the model or GLT specification syntax is correct.
Warning:
In summary.Anova.mlm(object$Anova, multivariate = FALSE) :
  one or more error SSP matrix:
corresponding non-sphericity tests and corrections not available"

I tried different solutions but no one worked (remove the covariates, control the dataTable file...).
Could it be due to the small sample size? What could be the problem?

Thanks in advance,

Bea

2

Hi Bea,

Regarding your use of 3dMVM:

I'm using 3dMVM to calculate the interaction effect of two within variables.

To properly model the interaction between the two within-subject variables, update:

-wsVars "Timing+HRFparam" \

to

-wsVars "Timing*HRFparam" \

I want to explore the effect of time separately on the three parameters of the HRF (time-to-peak, height, and fwhm).

Since these three parameters have distinct physical meanings and units, treating them as comparable levels of a within-subject factor may not be appropriate.

Your current GLT specifications:

...
      -gltLabel 2 'Pre_vs_Post_on_height' -gltCode 2 'Timing : 1*pre -1*post * HRFparam : 1*height' \
      -gltLabel 3 'Pre_vs_Post_on_timetopeak' -gltCode 3 'Timing : 1*pre -1*post * HRFparam : 1*time' \
      -gltLabel 4 'Pre_vs_Post_on_FWHM' -gltCode 4 'Timing : 1*pre -1*post * HRFparam : 1*fwhm' \

should be corrected to:

...
      -gltLabel 2 'Pre_vs_Post_on_height' -gltCode 2 'Timing : 1*pre -1*post  HRFparam : 1*height' \
      -gltLabel 3 'Pre_vs_Post_on_timetopeak' -gltCode 3 'Timing : 1*pre -1*post  HRFparam : 1*time' \
      -gltLabel 4 'Pre_vs_Post_on_FWHM' -gltCode 4 'Timing : 1*pre -1*post  HRFparam : 1*fwhm' \

Gang Chen

3

Dear Gang,

Thank you so much for your reply.

However, I tried to rerun the analyses with your suggestions, but unfortunately the same error came out.

How should I overcome this?

Thanks for any help

Bea

4

Bea,

Could you show the contingency table for Timing and HRFparam like the following?

...
                    HRFparam
Timing     height      time      fwhm
Pre           ?          ?         ?
Post          ?          ?         ?

Gang Chen