sdm-help-list > Group comparisons
Showing 1-3 of 3 posts
Results per page:
Jul 19, 2020  02:07 PM | Alex Rainer
Group comparisons
Hi everyone,

I have conducted 2 separately meta-analyses for 2 different populations (studies with young and and studies with old subjects). I noticed some significant differences in activation patterns between the 2 meta-analyses and I would like to quantify these. Is there anyway to conduct a meta analysis in which I can specific a contrast like Young > Old or Old > Young. I searched for the topic in this forum and from what I've gathered there may be a way to do this by using the Linear Model.

Since I'm not confident how this is done, I thought I'd check with those who are. Would it be correct to include 2 columns in the sdm_table.txt, 1 for old in which studies of old subjects have a value of 1 and studies of young subjects have a value of 0. The other column has the reverse coding. In the model specifications, I could choose 1 for young and -1 for old. This way, I'd have a contrast of Young > Old. I tried this and got this error:

gsl: ../../gsl-2.5/linalg/lu.c:266: ERROR: matrix is singular gsl: ../../gsl-2.5/linalg/lu.c:266: ERROR: matrix is singular Default GSL error handler invoked. Default GSL error handler invoked.

Could someone please let me know how to do this correctly?

Thank you in advance.

Jul 24, 2020  02:07 AM | Lydia Fortea - Instituto de Investigaciones Biom├ędicas August Pi i Sunyer (IDIBAPS)
RE: Group comparisons
Dear Alex,

Indeed, using the Linear Model function would be the proper approach.

However, instead of two columns as you did (you have the error because in the design matrix one colum is the opposite of the other, so is a singular matrix), you just need to put one column indicating if the patients are young (eg. 0) or old (eg 1). With this codification, for example, if you obtain a positive results would mean than OLD > YOUNG; whereas if you obtain a negative result, it will mean thant OLD < YOUNG.

Best regards, 
Jul 25, 2020  11:07 AM | Alex Rainer
RE: Group comparisons
Thank you for the answer, Lydia!