LOGIT

AFNI Message Board
1

I am trying to use the LOGIT function in 3dLME. We keep getting the following error:
Error in [.data.frame(lop$dataStr, , lop$vQV[1]) :
undefined columns selected
Calls: read.AFNI → parse.AFNI.name → [ → [.data.frame
Execution halted

The script we are using is below
3dLME -prefix respvnone_prevpost_anova_positive_regression
-model “time”
-qVars res
-ranEff ‘NA’
-LOGIT
-SS_type 3
-dataTable
Subj time res InputFile
s01 0 2.48 w_ssri6401sretc1sAUCidio+orig
s01 1 2.48 w_ssri8401sretc1sAUCidio+orig
s02 0 14.7 w_ssri6406sretc1sAUCidio+orig
s02 1 14.7 w_ssri8406sretc1sAUCidio+orig
s03 0 13.4 w_ssri6414sretc1sAUCidio+orig
s03 1 13.4 w_ssri8414sretc1sAUCidio+orig
s04 0 -1.35 w_ssri6458sretc1sAUCidio+orig
s04 1 -1.35 w_ssri8458sretc1sAUCidio+orig
s05 0 -1.77 w_ssri6479sretc1sAUCidio+orig
s05 1 -1.77 w_ssri8479sretc1sAUCidio+orig
s06 0 -1.68 w_ssri6480sretc1sAUCidio+orig
s06 1 -1.68 w_ssri8480sretc1sAUCidio+orig
s07 0 2.30 w_ssri6409sretc1sAUCidio+orig
s07 1 2.30 w_ssri8409sretc1sAUCidio+orig
s08 0 1.57 w_ssri6419sretc1sAUCidio+orig
s08 1 1.57 w_ssri8419sretc1sAUCidio+orig
s09 0 3.35 w_ssri6425sretc1sAUCidio+orig
s09 1 3.35 w_ssri8425sretc1sAUCidio+orig
s10 0 -2.57 w_ssri6433sretc1sAUCidio+orig
s10 1 -2.57 w_ssri8433sretc1sAUCidio+orig
s11 0 -7.64 w_ssri6434sretc1sAUCidio+orig
s11 1 -7.64 w_ssri8434sretc1sAUCidio+orig
s12 0 -3.52 w_ssri6435sretc1sAUCidio+orig
s12 1 -3.52 w_ssri8435sretc1sAUCidio+orig
s13 0 -2.38 w_ssri6439sretc1sAUCidio+orig
s13 1 -2.38 w_ssri8439sretc1sAUCidio+orig
s14 0 -5.45 w_ssri6441sretc1sAUCidio+orig
s14 1 -5.45 w_ssri8441sretc1sAUCidio+orig
s15 0 -3.72 w_ssri6444sretc1sAUCidio+orig
s15 1 -3.72 w_ssri8444sretc1sAUCidio+orig
s16 0 -0.54 w_ssri6456sretc1sAUCidio+orig
s16 1 -0.54 w_ssri8456sretc1sAUCidio+orig
s17 0 -2.39 w_ssri6488sretc1sAUCidio+orig
s17 1 -2.39 w_ssri8488sretc1sAUCidio+orig

This works when we remove the LOGIT function. We’ve gone down to the simplest iteration we could possible think of:

3dLME -prefix respvnone_prevpost_anova_positive_regression
-model “time”
-ranEff ‘NA’
-LOGIT
-SS_type 3
-dataTable
Subj time InputFile
s01 0 w_ssri6401sretc1sAUCidio+orig
s01 1 w_ssri8401sretc1sAUCidio+orig
s02 0 w_ssri6406sretc1sAUCidio+orig
s02 1 w_ssri8406sretc1sAUCidio+orig
s03 0 w_ssri6414sretc1sAUCidio+orig
s03 1 w_ssri8414sretc1sAUCidio+orig
s04 0 w_ssri6458sretc1sAUCidio+orig
s04 1 w_ssri8458sretc1sAUCidio+orig
s05 0 w_ssri6479sretc1sAUCidio+orig
s05 1 w_ssri8479sretc1sAUCidio+orig
s06 0 w_ssri6480sretc1sAUCidio+orig
s06 1 w_ssri8480sretc1sAUCidio+orig
s07 0 w_ssri6409sretc1sAUCidio+orig
s07 1 w_ssri8409sretc1sAUCidio+orig
s08 0 w_ssri6419sretc1sAUCidio+orig
s08 1 w_ssri8419sretc1sAUCidio+orig
s09 0 w_ssri6425sretc1sAUCidio+orig
s09 1 w_ssri8425sretc1sAUCidio+orig
s10 0 w_ssri6433sretc1sAUCidio+orig
s10 1 w_ssri8433sretc1sAUCidio+orig
s11 0 w_ssri6434sretc1sAUCidio+orig
s11 1 w_ssri8434sretc1sAUCidio+orig
s12 0 w_ssri6435sretc1sAUCidio+orig
s12 1 w_ssri8435sretc1sAUCidio+orig
s13 0 w_ssri6439sretc1sAUCidio+orig
s13 1 w_ssri8439sretc1sAUCidio+orig
s14 0 w_ssri6441sretc1sAUCidio+orig
s14 1 w_ssri8441sretc1sAUCidio+orig
s15 0 w_ssri6444sretc1sAUCidio+orig
s15 1 w_ssri8444sretc1sAUCidio+orig
s16 0 w_ssri6456sretc1sAUCidio+orig
s16 1 w_ssri8456sretc1sAUCidio+orig
s17 0 w_ssri6488sretc1sAUCidio+orig
s17 1 w_ssri8488sretc1sAUCidio+orig

and once again get the error when LOGIT is included but no error when it’s omitted. Any idea what we are doing wrong?

Thanks!

2

Could you explain under what circumstance you’re trying to run LOGIT?

3

We would like to do a nonlinear regression examining the relationship between brain activity pre and post treatment and residual symptoms. A linear regression did not produce results that survived cluster correction and the data look like the relationship may indeed be a logistic function. LOGIT within 3dLME is the only non-linear regression option we could find in the AFNI universe.

4

Currently the option -LOGIT in 3dLME is only for the scenario when the value for each subject is 0 or 1, as noted in the help: “The InputFile column is expected to list subjects’ responses in 0s and 1s.”

5

Is there any way to do a logistic regression in afni with quantitative variables?

6

Could you explain a little further exactly what you’re trying to model? How many explanatory variables? Which part involves LOGIT?

7

I want to explain the relationship between pre and post treatment whole brain activity and residual depressive symptoms.
We found 3dNLfim which looks more like what we want to do, we think the relationship is more of a drug response function and is not linear.
However, I am not sure how to incorporate the residual scores into this analysis.