Hi John,
the lowess plot can be misleading sometimes (as it is just the smoothed scatter data per group and not controlled for anything). It is not a regression analysis. Also note that the columns in your M matrix are not the same as in the table (because you usually drop the first 2 columns and then convert to numbers).
For setting up your design, you should talk to a biostatistician. I don’ know what variables are time-varying, or what the clinical scales is, do all subjects have a disease duration or only one group etc.
A design to test for different slopes per group (assuming group is 0 and 1) would be: 1-column time group timeXgroup (more columns with covars such as age sex etc) the 4th column can be used for testing a group-slope interaction (if yearly thickness rates differ across groups).
If you assume slope does not differ across groups, but rather depends on disease duration, you can replace group with disease duration (and you can add a group column also for allowing an offset per group, or you can include the groupXtime interaction and even add a groupXtimeXdiseasedurtation column). You see that it can get complicated if you like. The next question will be what of these should be random effects. I would recommend to only have the intercept (1-column) as random, and maybe the time column, but even that may not be necessary. There is a description on the wiki how to compare models to determine if including the time random effect makes sense or not.
Best, Martin
On Jan 13, 2016, at 3:07 PM, John Anderson j.anderson@publicist.com wrote:
Dear Dr Martin, I have longitudinal database for subjects scanned multiple times. I want to study the changes in cortical thickness over time. I prepared my longitudinal table ( column#1 fsid column#2 fsid-base column#3 group column#4 bilateral motor cortex thickness column#5 time (years) column#6 disease duration) Then I followed the instructions like in wiki... and I plot the changes in motor cortex over time using the command :
lme_lowessPlot(M:,5),Y(:,1)m0.7,M(:,3))
My questions are:
- How can I regress the effect of disease duration (column#6) ? is it by running the previous command line by replaceing M(:,5) ( time) by M(:,6) disease duration?
- If I am interested in studying the change in cortical thickness by the change in clinical scale over time is it correct to replce the coloumn #5 ( time) by a column for the clinical scales?
Thanks in advance for any advice!
Bests, John _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer