Dear SPMers,

I encountered the global normalisation problem in Basic Models as Eraldo and
Anna did. The below is the posts they made to the list in the past.


Eraldo Paulesu wrote:

I would like to perform a one-way factorial ANOVA on resting state PET scans
in a sample where I have 4 groups (one scan for each subject)

Apparently the ANOVA in simple models of SPM99 does not
allow one to correct for global counts (I would like proportional scaling here)
or, at least, the SPM.ps page reports - no correction for global counts.

http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind01&L=SPM&P=R484143&I=-3


Anna Barnes wrote:

I just ran a Basic models paired t-test analysis and entered no grand  
mean scaling and global normalisation as 0.8 of mean but when the  
design matrix came up it stated that I had selected no global  
normalisation. I ran it twice to make sure.


http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0510&L=SPM&P=R36608&I=-3&m=13895


I would like to perform a one-way ANOVA on SPECT scans in a sample of which I
have three groups (one scan for each subject). With one-way ANOVA in Basic
models, even though it accepts global normalisation on GUI, design matrix states
I had selected no global normalisation. In other studies, I tried to use the
paired t-test or simple regression of Basic models, and same things happened in
these models. However, PET models handled global normalisation properly. So I
want to know what the difference lies between Basic models and PET models.
Design matrix should be the same except for global normalisation...

I'm grateful for any suggestion.

Best Regards,

Kiyotaka

----------------
Kiyotaka NEMOTO, MD
Dept. of Psychiatry
Ibaraki Prefectural Tomobe Hospital
654 Asahicho Nishiibaraki, Ibaraki
309-1717 Japan
E-mail: [log in to unmask]