Email discussion lists for the UK Education and Research communities

## SPM@JISCMAIL.AC.UK

#### View:

 Message: [ First | Previous | Next | Last ] By Topic: [ First | Previous | Next | Last ] By Author: [ First | Previous | Next | Last ] Font: Proportional Font
 LISTSERV Archives SPM Home SPM May 2011

#### Options

Subject:

Re: VBM stat with 2 groups

From:

Jonathan Peelle <[log in to unmask]>

Reply-To:

Jonathan Peelle <[log in to unmask]>

Date:

Thu, 5 May 2011 12:44:24 -0400

Content-Type:

text/plain

Parts/Attachments:

 text/plain (61 lines)
 ```Dear Soufiane, > I have 2 groups of 15 subjects for a VBM study and made the preprocessing. > Now I'd like to do the stats and I'm a little bit lost to covariate with age and sex. > How to put the vectors for sex (-1; 1?) and  age (difference with the mean?) > Is the order of the value the order of the subjects successively in the 2 groups (the 16th value is the one for the 1st subject of the 2d group for example)? It sounds as if you want to compare tissue volume in two groups while controlling for age and sex differences. To start with, let's say you just want to control for age differences. In this case you would specify a 2-sample t-test and add a covariate for age; this will give you 3 columns in your design matrix (group 1, group 2, age). In this case, as with any covariate, you need one value per image, and these values are in the same order as the images you select. So in this case, you've selected 30 images (15 group 1, 15 group 2). You are exactly right that the 16th value in the age covariate should be the age for the 16th image = first subject in group 2. You will want to mean center the covariate (which is the default in SPM8). To test the difference between group 1 and group 2, you could then do a t-test of: [1 -1 0] (group 1 > group 2) or [-1 1 0] (group 2 > group 1) You could look at the effect of age with [0 0 1] or [0 0 -1]. In the case where you want to control for sex differences, one common approach is to model males and females separately within each group, which would require a slightly different model. However, since you only have 15 subjects per group, I don't know that this will be particularly productive. A general comment with regard to covariates is that if your groups are fairly evenly matched, then including an additional covariate will reduce the error (which is good) but probably not change the overall result. If your groups are not particularly well balanced, then your group difference is confounded with these other factors, and including them as covariates cannot undo this. Hope this helps! Best regards, Jonathan -- Dr. Jonathan Peelle Department of Neurology University of Pennsylvania 3 West Gates 3400 Spruce Street Philadelphia, PA 19104 USA http://jonathanpeelle.net/ ```

#### RSS Feeds and Sharing

JiscMail is a Jisc service.

View our service policies at https://www.jiscmail.ac.uk/policyandsecurity/ and Jisc's privacy policy at https://www.jisc.ac.uk/website/privacy-notice

For help and support help@jisc.ac.uk