Hi Ben
I think I see what has happened:
It looks like column 2 is your age regressor.
Column 1 looks like a column of 1's. My bet is you chose the 1 sample t-test
design and entered a covariate of age. This contrast is essentially saying
show me the areas in the image that are different from 0. This will be
everywhere as you have discovered except there are probably some holes in
the middle where there is no gray matter.
Instead click on the line that says design in the batch box (I assume you
have chosen factorial design) and choose the multiple regression design.
Then enter your scans and the covariate (choice right under the scans line).
Your design will now consist of a single column (make sure you have mean
corrected your age covariate). Now you can run a contrast on age.
You may want to consider including other covariates in your design such as
total gray matter or total intracranial volume to account for these or other
nuisance effects. These design issues are covered in other email messages or
in the SPM documentation or you can look at Christian Gaser's wonderful VBM
site.
hth,
darren
----------
> -----Original Message-----
> From: SPM (Statistical Parametric Mapping)
> [mailto:[log in to unmask]] On Behalf Of Benjamin Kelemen
> Sent: Friday, November 16, 2007 3:17 PM
> To: [log in to unmask]
> Subject: [SPM] Correlation analysis problems
>
> Hello SPMers,
>
> I am still learning to use SPM5 for a VBM analysis. I have
> had a few problems with my analysis.
>
> The first problem I have had was in my attempt to do a
> correlation analysis of age and gray matter volume. However,
> in my correlation analysis, the entire brain has consistently
> been showing up as significant for a positive correlation and
> nothing has shown up significant for a negative correlation
> (as shown in the attachment). This doesn't seem right.
> However, I have gotten this result so consistently that I
> assume I must be doing some step wrong. I have tried this
> with both the unmodulated and the modulated images. I have
> also checked the values of each image to make sure that there
> wasn't one outlier that was distorting all results.
>
> I have also had a number of significant blobs on the outside
> of the brain.
> However, that problem was helped by using an absolute
> threshold or an explicit brainmask.
>
> Thanks so much for the help. You guys are great.
>
> Ben
>
|