The second option only works for 2 conditions (or 2 measures per
subject). If you have missing data, then you can't do much in the case
of only 2 measurements.
Repeated-measures ANCOVAs don't have a uniform solution across
statistical packages, thus I would avoid them.
I don't currently have any solutions for repeated-measures ANCOVAs or
correlations across repeated-measures beyond the simple paired case.
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
Website: http://www.martinos.org/~mclaren
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 Fri, Apr 13, 2012 at 2:41 PM, Arman Eshaghi <[log in to unmask]> wrote:
>
> Sorry for hijacking this thread, just out of curiosity what would you do for
> the second option in this setting? (correlation of change in data with
> covariate of interest in repeated measure mixed-effects models -to account
> for missing time-points in repeated measure design-)
>
> All the best,
> Arman
>
>
> On Fri, Apr 13, 2012 at 10:18 PM, MCLAREN, Donald <[log in to unmask]>
> wrote:
>>
>> Just an added note on the topic. It is statistically invalid to test
>> between-subject effects with the within-subject error terms produced
>> by FSL and SPM. GLM flex will handle this properly, but prohibits
>> repeated-measures ANCOVAs (such as your model - as there isn't a
>> consistent way to compute rm-ANCOVAs across programs).
>>
>> The immediate question that comes to mind is what is your hypothesis,
>> are you trying to:
>> (a) simple correlation of everything [e.g. correlation as if there was
>> a single group];
>> (b) correlation of change in data with your covariate [e.g.
>> correlation with the difference images];
>> (c) the difference in correlation at baseline and 5 years?
>>
>> 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
>> Website: http://www.martinos.org/~mclaren
>> 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 Fri, Apr 13, 2012 at 8:27 AM, Irwin, William <[log in to unmask]>
>> wrote:
>> > Singe-
>> >
>> > You cannot treat the image data at each time point-- 2 per patient-- and
>> > independent measures.
>> >
>> > It is a "repeated measures" issue, and randomization does not account
>> > for the dependence of the measures (as some might claim).
>> >
>> > The most straight forward approach would be a repeated-measures (or
>> > equivalent regression) ANOVA analysis.
>> >
>> > If you are correlating image data at time 1, time 2 and some 3rd
>> > measure, there are methods to determine if the r(1 w/ measure) and r(2
>> > w/measure) are reliably different.
>> >
>> > A very good treatment and general references for this is "Biostatistical
>> > Analyses" (5th edition) by J. Zar.
>> >
>> > Regards,
>> > William
>> >
>> > |-----Original Message-----
>> > |From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On
>> > Behalf
>> > |Of Signe Jeppesen
>> > |Sent: Friday, April 13, 2012 12:44 AM
>> > |To: [log in to unmask]
>> > |Subject: [FSL] Correlation analysis on repeated measures
>> > |
>> > |Hi FSL
>> > |
>> > |I am trying to do a correlation analysis on my dataset of patients that
>> > have
>> > |been scanned twice with 5 years in between.
>> > |
>> > |I would like to do a correlation analysis of the data and my EVīs but
>> > still
>> > |account for the repeated measures.
>> > |
>> > |So far I have done a correlation analysis in randomise including a
>> > group file
>> > |(.grp ) that links the 2 scans of the patient together.
>> > |Is this the right way to do it?
>> > |
>> > |Or should I not do a repeated measure analysis and just treat every
>> > scan as a
>> > |different patient in my correlation analysis?
>> > |
>> > |Thank you
>> > |Signe
>
>
|
