3dLME Model Error / Script Crashing

AFNI Message Board
1

Hi there,

I’m currently trying to run a simple 3dLME with the following:

3dLME -prefix final_PCLTotal_LME -jobs 24
-model “Time*PCLTotal+age+Gender”
-qVars “age”
-ranEff “~1+age”
-SS_type 3
-dataTable @CONN_LME_DataTable.txt

The script successfully reads the input files, but then gets stuck on the model test before killing it after about 5-10 minutes (it does this on its own, I do not end it). I have run file_tool on my script and run the FIXED version, so I am guessing there is a problem with the model since I don’t have any GLT coding in the script. I’m not sure what to change about the model, but any advice would be helpful.

Here’s the output I get:

Reading input files now…
Reading input files: Done!

If the program hangs here for more than, for example, half an hour,
kill the process because the model specification or the GLT coding
is likely inappropriate.

Killed

2

-model “Time*PCLTotal+age+Gender”
-qVars “age” \

Is “age” the only quantitative predictor? Make sure that all the quantitative predictors are declared.

3

You’re right, PCLTotal should also be included in the qVars line. Time and Gender are both dichotomous variables, so they would not go in the qVars, correct? Here is my model now:

3dLME -prefix final_PCLTotal_LME -jobs 24
-model “Time*PCLTotal+age+Gender”
-qVars “PCLTotal,age”
-ranEff “~1+age”
-SS_type 3
-dataTable @CONN_LME_DataTable.txt

It doesn’t kill the script now, but the model is still failing the model test. Is there anything else I should be changing?

4

Does ‘PCLTotal’ vary between the two ‘Time’ points? If not, the model

-model “Time*PCLTotal+age+Gender” \

does not make sense, and you can try

-model “Time+PCLTotal+age+Gender” \

5

it does, PCLTotal should be a different value for each time point. Here is a clip of my data table.
Note: I had changed the variable names in the original post for simplicity’s sake, so that is not the issue. Also there are 121 subjects, but this is just a portion:

Subj Time Gender Racial_Category age2wk PCLTotal_Both Input.File
2 1 1 4 24.8 35 REST/results/firstlevel/allsubs_ICC/BETA_Subject002_Condition001_Measure002_Component001.nii
2 2 1 4 24.8 24 REST/results/firstlevel/allsubs_ICC/BETA_Subject002_Condition003_Measure002_Component001.nii
3 1 1 5 20.9 51 REST/results/firstlevel/allsubs_ICC/BETA_Subject003_Condition001_Measure002_Component001.nii
3 2 1 5 20.9 46 REST/results/firstlevel/allsubs_ICC/BETA_Subject003_Condition003_Measure002_Component001.nii
5 1 0 4 25.6 49 REST/results/firstlevel/allsubs_ICC/BETA_Subject005_Condition001_Measure002_Component001.nii
5 2 0 4 25.6 37 REST/results/firstlevel/allsubs_ICC/BETA_Subject005_Condition003_Measure002_Component001.nii \

I am curious, is the issue that I am trying to use the beta maps from CONN as the input files? I had originally ran this ICC (intrinsic connectivity contrast) in CONN with the intention of completing the LME in AFNI; however, maybe this is messing with the model.

6

is the issue that I am trying to use the beta maps from CONN as the input files?

No, almost all AFNI programs accept NIfFI files as input.

Try changing the column name “Input.File” to “InputFile”

7

Ah, sorry I copied this from my excel spreadsheet. In the text file I had changed it to InputFile already, but it still fails the model.

8

I don’t have the full information about your data structure, so I’d have to keep guessing:

Does ‘age’ vary within subject? If not, specification

-ranEff “~1+age” \

does not make sense. Change it to

-ranEff “~1” \

or

-ranEff “~1+PCLTotal” \

9

Sorry for all the back and forth, let me know if there’s anything I can send to make things clearer.
Age does not vary within subjects, so I tried “~1” but it failed. I also tried the “~PCLTotal” which also failed, although PCLTotal is the primary outcome measure I am interested in. Essentially, I have two sessions, one at 2 weeks and one at 6 months, and I want to assess if changes in connectivity between the 2 week and 6 month sessions relates to changes in the PCLTotal score (a self-administered symptom scale). Age and Gender are included as covariates. Does any of that help?