Dear Donald and list,
I agree that the correct way to model within-subjects effects is a flexible factorial design with a subject factor, assuming equal variance and non-sphericity over conditions. I have, however, seen publications of within-subjects effects where the subject terms were modeled in the form of covariates in a full factorial design. What are your thoughts on the validity of this approach compared to the subject factor in a flexible factorial design.
Best wishes,
Martin Dietz
[log in to unmask]
________________________________________
From: SPM (Statistical Parametric Mapping) [[log in to unmask]] On Behalf Of MCLAREN, Donald [[log in to unmask]]
Sent: 02 June 2011 05:42
To: [log in to unmask]
Subject: Re: [SPM] second level analysis affected by subject order
Dear All,
I was able to replicate the problem as well as solve the problem. If you are doing a repeated measures ANOVA, paired t-test, etc. then INDEPENDENCE should be set to No and VARIANCE set to Equal. In doing so, you will get the same values irregardless of subject order. If you choose Unequal variance, then subject order matters; of course, since the scans come from the same people, the variance should be the same. If you choose Yes for Independence, then you are treating the scans as coming from different subjects and that is statistically incorrect.
Personally, and I'll be presenting at OHBM, the flexible factorial should be used for within-subject designs. In using the flexible factorial, you explicitly model the relationship between scans by including a subject term.
Best Regards, Donald McLaren
=================
D.G. McLaren, Ph.D.
Postdoctoral Research Fellow, GRECC, Bedford VA
Research Fellow, Department of Neurology, Massachusetts General Hospital and Harvard Medical School
Office: (773) 406-2464
=====================
This e-mail contains CONFIDENTIAL INFORMATION which may contain PROTECTED HEALTHCARE INFORMATION and may also be LEGALLY PRIVILEGED and which is intended only for the use of the individual or entity named above. If the reader of the e-mail is not the intended recipient or the employee or agent responsible for delivering it to the intended recipient, you are hereby notified that you are in possession of confidential and privileged information. Any unauthorized use, disclosure, copying or the taking of any action in reliance on the contents of this information is strictly prohibited and may be unlawful. If you have received this e-mail unintentionally, please immediately notify the sender via telephone at (773) 406-2464 or email.
On Tue, May 31, 2011 at 5:07 PM, David Soto <[log in to unmask]<mailto:[log in to unmask]>> wrote:
Hi, yes, say you have 3 subjects and one factor with 2 levels
at the highest level analyses you may specify the betas in different orders
subj1 beta1 subj1beta2
subj2 beta1 subj2beta2
subj3 beta1 subj3beta2
or
subj3 beta1 subj3beta2
subj2 beta1 subj2beta2
subj1 beta1 subj1beta2
or any other combination
you might see a similar set of clusters in both cases but order influences the stats in a funny way
guess there may be implications to assess significance when the effects are borderline
but in any case it is weird and am curious to learn what is the reason of this behaviour in the first place
rgds.
|