Thanks Steve. I think we'd want to contrast the behavioral correlation
between the two groups. So, are you saying that in the 2nd GLM I should
have 2 groups and 4 EVs, with EV1 and EV2 being exactly the same as I
did in GLM 1 and then EV3 and EV4 would be the demeaned scores padded
with zeros? Also, I don't actually know how to demean scores? Is there
an FSL command that I can use to do this or is it something I need to do
with a different program? Thanks again.
-Andy
-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
Behalf Of Steve Smith
Sent: Tuesday, December 05, 2006 1:47 PM
To: [log in to unmask]
Subject: Re: [FSL] GLM GUI setup
Hi,
On 5 Dec 2006, at 16:16, Andrew Silver wrote:
> I am trying to figure out how to setup randomise to do a correlation
> between FA values and a behavioral measure. Our setup has a patient
> group and a control group. If I'm reading the other questions
> correctly I believe I should do 2 GLM_GUIs, with the first making the
> design.mat file and the design.con file, and the second making the
> .mat file for the -x flag. First off, is this the correct setup?
Yes.
> When I made the first
> GLM_GUI I had 2 groups and 2 EVs. For each person in Group 1 I had
> EV1=1
> and EV2=0, and for Group 2 I had EV1=0 and EV2=1. I set the contrast
> to be 1 -1. Is this the correct setup for my first GLM?
Yes.
> As for the second
> GLM, I am confused on how to set this up. Should I have both
> groups in
> it? How many EVs should I have? Do I have 2 groups with 1 EV and
> have
> that EV column filled with their scores on the behavioral test?
> When I
> tried that I was told that only 1 group could have non-zero EV
> values. I
> am also confused as to whether I should have actual behavioral
> scores in
> the EV column or if they should be demeaned. Any help would be much
> appreciated. Thanks.
Mostly this all depends on whether you want to test whether all
subjects in both groups correlate with the behavioural measure on
average (in which case use a single EV with demeaned behavioural
scores for all subjects and put the group membership EVs into the -x
confoudn matrix) or whether you want to contrast the behvaioural
correlation between the two groups (in which case the confound matrix
is the same as above, and the "real" model has two EVs, containing
the different group's behavioural scores, each demeaned before
padding with zeros; the contrast is then [1 -1] etc.).
Cheers.
> -Andy
------------------------------------------------------------------------
---
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
------------------------------------------------------------------------
---
|