help
help > RE: CONN exporting variables to first and second level
May 24, 2018 09:05 AM | Alfonso Nieto-Castanon - Boston University
RE: CONN exporting variables to first and second level
Dear Joga,
The quality control covariates created by CONN's preprocessing and/or denoising steps are NOT automatically taken into account in any analyses unless you explicitly specify so. Typically you would want to use the "histogram of QC subject-level values" plot in the QA gui to evaluate the range of subject-motion and valid vs. outlier scans in your data, and identify potential outlier subjects if any. You may also use the "Tools.Calculator" gui to evaluate potential differences in subject-motion or other QC covariates among subject groups or conditions. If there are residual effects that you want to control for (e.g. differences in subject-motion between groups) you may then decide to do so when you run second-level analyses simply by including one or more of the QC* covariates as additional covariates of no interest in your analyses (e.g. ANCOVA test looking at between-group differences in connectivity when controlling for potential differences in average-subject-motion between the groups).
Regarding the second-level warning messages and drop from 12 to 4 subjects, my first guess would be that the "AllSubjects" covariate may have been manually modified somehow from the "all 1's" default values so that it may include now some NaN values (missing-data label; that could cause the drop from 12 to 4 subjects in your analyses as missing-data subjects will be removed from the analysis) and/or values other than 1's (that could also explain the "possibly incorrect model: no constant term modeled" warning message). If in doubt just send me your conn_*.mat file and I will take a quick look to see if I can diagnose the cause of this errors.
Last, to perform a paired t-test comparing your two conditions (pre- and post- treatment), simply select:
in "subject-effects" list: AllSubjects
in "between-subjects contrast": 1
in "conditions": pre- and post-
in "between-conditions contrast": -1 1
(and just to double-check note that the results should report T(11) statistics/degrees of freedom)
Best
Alfonso
Originally posted by Joga Chaganti:
The quality control covariates created by CONN's preprocessing and/or denoising steps are NOT automatically taken into account in any analyses unless you explicitly specify so. Typically you would want to use the "histogram of QC subject-level values" plot in the QA gui to evaluate the range of subject-motion and valid vs. outlier scans in your data, and identify potential outlier subjects if any. You may also use the "Tools.Calculator" gui to evaluate potential differences in subject-motion or other QC covariates among subject groups or conditions. If there are residual effects that you want to control for (e.g. differences in subject-motion between groups) you may then decide to do so when you run second-level analyses simply by including one or more of the QC* covariates as additional covariates of no interest in your analyses (e.g. ANCOVA test looking at between-group differences in connectivity when controlling for potential differences in average-subject-motion between the groups).
Regarding the second-level warning messages and drop from 12 to 4 subjects, my first guess would be that the "AllSubjects" covariate may have been manually modified somehow from the "all 1's" default values so that it may include now some NaN values (missing-data label; that could cause the drop from 12 to 4 subjects in your analyses as missing-data subjects will be removed from the analysis) and/or values other than 1's (that could also explain the "possibly incorrect model: no constant term modeled" warning message). If in doubt just send me your conn_*.mat file and I will take a quick look to see if I can diagnose the cause of this errors.
Last, to perform a paired t-test comparing your two conditions (pre- and post- treatment), simply select:
in "subject-effects" list: AllSubjects
in "between-subjects contrast": 1
in "conditions": pre- and post-
in "between-conditions contrast": -1 1
(and just to double-check note that the results should report T(11) statistics/degrees of freedom)
Best
Alfonso
Originally posted by Joga Chaganti:
Dear Stephen and alfonso
It was a pity i could not stay through the conference and hence have to bother you all with these questions. I was hoping if you could help me with this advise at the first level and second level co variates
The cohort of my research is 12 patients pre and post treatment and 8 controls.
I initially looking at the differences in connectivity in the patients group only. i have finished the demonising step and assume that the deposing step created covariates automatically are taken by CONN to factor them in to first level analysis. IS THIS ASSUMPTION CORRECT??
In the GUI for second level the Between subject contrast and between the condition contrast and between the sources contrast how do we plan for my cohort. i assumed that it should be 1; -1 0; -1 and leave the third one blanc. But regardless of how i try to change the GUI does not allow me to change and the numbers go back to default. and error message flashes in on the design.
Its interesting that the patients numbers are 12 ,up to first level and once we put them in to second level the number drops to 4 and i do not understand that as well.
If i bring the controls in to factor the comparison of connectivity between the patients and controls at both the time points how should i plan the GLM ?
i am attaching the error messages all here.
Please help
reagrds
Joga
It was a pity i could not stay through the conference and hence have to bother you all with these questions. I was hoping if you could help me with this advise at the first level and second level co variates
The cohort of my research is 12 patients pre and post treatment and 8 controls.
I initially looking at the differences in connectivity in the patients group only. i have finished the demonising step and assume that the deposing step created covariates automatically are taken by CONN to factor them in to first level analysis. IS THIS ASSUMPTION CORRECT??
In the GUI for second level the Between subject contrast and between the condition contrast and between the sources contrast how do we plan for my cohort. i assumed that it should be 1; -1 0; -1 and leave the third one blanc. But regardless of how i try to change the GUI does not allow me to change and the numbers go back to default. and error message flashes in on the design.
Its interesting that the patients numbers are 12 ,up to first level and once we put them in to second level the number drops to 4 and i do not understand that as well.
If i bring the controls in to factor the comparison of connectivity between the patients and controls at both the time points how should i plan the GLM ?
i am attaching the error messages all here.
Please help
reagrds
Joga
Threaded View
| Title | Author | Date |
|---|---|---|
| Joga Chaganti | May 19, 2018 | |
| Alfonso Nieto-Castanon | May 24, 2018 | |
| Joga Chaganti | May 25, 2018 | |