Dear Danny,
Whether or not there is a constant in the design matrix makes no
difference to the result when you do a 1 -1 contrast ie. to test the
difference between groups.
So there's basically no difference between the approaches.
However, without a constant you will also be able to test for the effect
of each group alone, ie 1 0 and 0 1. You cannot do this if there
is a constant in the design matrix, as each of these columns can be
gotten from a linear combination of the other column plus the constant
column (ie. they are `colinear').
You may be interested to know that in SPM8 you can specify a design from
the GUI, then save that specification as a script. You can then edit
that script file to enter eg. info from another subject or other subtle
change in design.
Best,
Will.
Qiang (Danny) Chen wrote:
>
>
> Dear Dr. Penny:
>
>
>
> I have a question about Two sample T test. I cannot find answer from SPM
> email list, so I have to bother you directly.
>
>
>
> I have been using spm_spm_ui.m for my batch scripts since spm2. It seems
> spm5 inherited this script, but spm5 interface uses
> spm_config_factorial_design.m, not spm_spm_ui.m. The difference between
> these two scripts on Two-sample T test is, previous one has constant,
> but interface one doesn’t have. I am curious to know what difference on
> modeling between these two scripts is, and if they can cause any big
> difference.
>
>
>
> Thanks,
>
>
>
> */Qiang (Danny) Chen, Ph.D./*
>
>
>
> Clinical Brain Disorders Branch (CBDB)
> Genes, Cognition and Psychosis Program (GCAP)
> NIMH, NIH
>
>
>
--
William D. Penny
Wellcome Trust Centre for Neuroimaging
University College London
12 Queen Square
London WC1N 3BG
Tel: 020 7833 7475
FAX: 020 7813 1420
Email: [log in to unmask]
URL: http://www.fil.ion.ucl.ac.uk/~wpenny/
|