Hi Jonathan,
I would recommend to run all subjects to the longitudinal stream only once and include as many time points for each subject as possible. That way you have only one measure per time point. Then I would use a linear mixed effects model (which can deal with missing data) and analyze the results. We have matlab tools for the LME analysis, but you can easily use asegstats2table to stack all measures into a table and import it to SPSS or whatever.
Best, Martin
On 19 Jan 2017, at 16:15, Greenberg, Jonathan JGREENBERG5@mgh.harvard.edu wrote:
Dear FreeSurfer users
I conducted a longitudinal study with 3 timepoints: baseline, 8 weeks and 24 weeks. Between baseline and 8 weeks participants were randomized to one of 3 training programs. I want to assess changes in gray matter volume from baseline to 8 weeks, and baseline to 24 weeks, and also determine how these changes correlate with various behavioral outcome measures. The correlations will be conducted in SPSS, so I need to export values from each time point to enter into the regression matrix.
The issue is that about 20 people dropped out after the 8 week assessment. So to just ask the question about changes in brain structure I created two different processing streams, one for all 75 subjects who have pre and 8 week data, and a second one for the 58 individuals who have pre and 24 week data. Therefore, participants who have data for all 3 time points had their pre and 8-week data analyzed in both processing streams. These separate processing streams yield different pre values for these individuals (e.g. the hippocampal volume of participant X at “pre” in the first processing stream is different than the pre hippocampal volume of the same participant in the second processing stream).
My question is about the proper way to combine data from these two processing streams so that I have one Excel column for each of the 3 time points to be analyzed in SPSS with all participants included. One approach would be to use the values from the second processing stream (which includes the 24 week time point) for the participants who have all 3 time points, and use the data from the first processing stream only for those who are missing the 24 week time point. But I am not sure whether this is acceptable given the discrepancy in values between the two processing streams. So, is there better way to combine the data from both streams into these 3 time-point columns?
Thank you
Jonathan
Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu mailto:Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer