3dLME glfCode

AFNI Message Board
1

AFNI version info (Precompiled binary linux_centos_7_64: Jun 30 2020 (Version AFNI_20.1.18 'Otho'):
Hi,
I'm trying to run 3dLME (3 x 3 design) where I have 3 groups and 3 timepoints. I want to run a glf test where I only run a group analysis at baseline. This is how I did it.
3dLME finishes without any problem, but when I go check the results for the glf test, I get a chisq instead of F. Just checking that those results are what they should be.

Thanks!

3dLME -prefix path/results_3dLME_SN_all_groups.nii -jobs 12 \
-mask path/SN_mask.nii \
-model 'group*session+age+sex+FD' \
-qVars 'age,FD' \
-ranEff '~1+FD' \
-SS_type 3 \
-num_glf 1 \
-glfLabel 1 'baseline' -glfCode 1 'session : 1*ses-01 group : 1*HC -1*Affective & 1*HC -1*NonAffective' \
-dataTable \
Subj group session age sex FD InputFile \
sub-control01	HC	ses-01	28	2	0.1217738	/path/connmap_sub-control01_ses-01_SN.nii \
sub-control01	HC	ses-02	28	2	0.1749546	/path/connmap_sub-control01_ses-02_SN.nii \
sub-control01	HC	ses-03	28	2	0.150672	/path/connmap_sub-control01_ses-03_SN.nii \
sub-patient06	Affective	ses-01	24	1	0.106127	/path/connmap_sub-patient06_ses-01_SN.nii \
sub-patient06	Affective	ses-02	24	1	0.1434903	/path/connmap_sub-patient06_ses-02_SN.nii \
sub-patient06	Affective	ses-03	24	1	0.149757	/path/connmap_sub-patient06_ses-03_SN.nii \
sub-patient01	NonAffective	ses-01	25	1	0.1971614	/path/connmap_sub-patient01_ses-01_SN.nii \
sub-patient01	NonAffective	ses-02	25	1	0.2590447	/path/connmap_sub-patient01_ses-02_SN.nii \
sub-patient01	NonAffective	ses-03	25	1	0.254582	/path/connmap_sub-patient01_ses-03_SN.nii \
2

when I go check the results for the glf test, I get a chisq instead of F.

Conventional ANOVA employs F-statistics due to the clear definition of degrees of freedom. However, in hierarchical models, accurately specifying degrees of freedom becomes challenging, leading to the use of chi-square statistics as an alternative.

Consider properly centering the two quantitative variables age and FD as discussed here. Additionally, consider using 3dLMEr instead of 3dLME as the latter is gradually being phased out.

Gang