Hi Knut

There are a few confusing things in your question post. First, it seems that your attached images are depicting p-values. We usually use -log10(pvalue) format for visualization in tksurfer. The sig.mgh file saved by lme should be in that format. However we didn't use tksurfer to build the figures for the paper you mentioned. We instead just used matlab-based geometric objects and figures. Second, for mass-univariate analyses the contrast matrix is given to lme as a matlab structure CM.C = [0 0 1 0 0 0] not a plain matrix C. Perhaps you should try first the vertex-wise lme model to ensure that all your parameters (design matrix, data matrix, contrast matrix, etc...) are right and then give it a try to the more complex spatio-temporal lme model using the same parameters.

eg. lhstats = lme_mass_fit_vw(X,[1 2],Y,ni,lhcortex);

Then you can compare the results to have a better idea about what is going on. 

Best
-Jorge


El Jueves 5 de diciembre de 2013 13:29, Martin Reuter <mreuter@nmr.mgh.harvard.edu> escribió:
Hi Jorge,

this was on the freesurfer list. Do you know what is going on ?

Best, Martin


-------- Original Message --------
Subject: [Freesurfer] Problems with longitudinal analysis
Date: Wed, 4 Dec 2013 22:04:07 +0100
From: Knut J Bjuland <knutjorgen@outlook.com>
To: freesurfer <freesurfer@nmr.mgh.harvard.edu>


Hi,

I have used a linear mixed model for longitudinal data analysis on a 
FreeSurfer 5.3, Matlab 2013b on a system with Ubuntu 12.04.

I used mris_preproc --qdec-long qdec.table.dat --target fsaverage --hemi 
lh --meas thickness --out lh.thickness.mgh to concatenate thickness surf 
files, and smoothed with mri_surf2surf --hemi lh --s fsaverage --sval 
lh.thickness.mgh --tval lh.thickness_sm30.mgh --fwhm-trg 30 --cortex 
–noreshape.

I then used the same command and options in a Matlab script, as shown at 
http://freesurfer.net/fswiki/LinearMixedEffectsModel found in the 
example for mass-univariate data analyses.

.

I used this design matrix: interception, time, group, group*time, 
gender, gender*time, and I used two random effect on interception and 
time. I did not compare the two random effect models with one random 
effect model.

I used this contrast vector C=[0 0 1 0 0] for these images which are 
included in the email.

The script ran fine but when I looked at sig.mgh, it contained many 
visible quadrate shapes boxes even after FDR correction, and the
image looks quite different than the image in the paper in Neuroimage 
(Spatiotemporal Linear Mixed Effects Modeling for the Mass-univariate 
Analysis of Longitudinal ).


Any idea what went wrong here?


Thank you!

Best regards,
Knut Jørgen


The information in this e-mail is intended only for the person to whom it is
addressed. If you believe this e-mail was sent to you in error and the e-mail
contains patient information, please contact the Partners Compliance HelpLine at
http://www.partners.org/complianceline . If the e-mail was sent to you in error
but does not contain patient information, please contact the sender and properly
dispose of the e-mail.