Hi,
I'm trying to set up a glm design matrix for a 2x4 anova with 1 repeated measure. I have a control group and a patient group with 4 timepoints for each subject. There is some discussion in previous postings about how to set up the model when you have more than 2 factors and multiple levels, but no examples. Could anyone who has set up such a higher order model provide an example (eg 2x3, 3x3, 2x4 etc)?
In trying to set up this model, I have come up with only 2 designs aren't rank deficient. One is similar to the tripled ttest, though I don't think it is correct. The other tries to model the individual cells (but not for timepoint 4) and is outlined below.
Can someone tell me if this is correct or help me correct this? This models 2 control subjects and 3 patients.
thanks,
Kyle Kern
cont1 T1 1 0 0 0 0 0 1 0 0 0 0
cont1 T2 0 1 0 0 0 0 1 0 0 0 0
cont1 T3 0 0 1 0 0 0 1 0 0 0 0
cont1 T4 0 0 0 0 0 0 1 0 0 0 0
cont2 T1 1 0 0 0 0 0 0 1 0 0 0
cont2 T2 0 1 0 0 0 0 0 1 0 0 0
cont2 T3 0 0 1 0 0 0 0 1 0 0 0
cont2 T4 0 0 0 0 0 0 0 1 0 0 0
patn1 T1 0 0 0 1 0 0 0 0 1 0 0
patn1 T2 0 0 0 0 1 0 0 0 1 0 0
patn1 T3 0 0 0 0 0 1 0 0 1 0 0
patn1 T4 0 0 0 0 0 0 0 0 1 0 0
patn2 T1 0 0 0 1 0 0 0 0 0 1 0
patn2 T2 0 0 0 0 1 0 0 0 0 1 0
patn2 T3 0 0 0 0 0 1 0 0 0 1 0
patn2 T4 0 0 0 0 0 0 0 0 0 1 0
patn3 T1 0 0 0 1 0 0 0 0 0 0 1
patn3 T2 0 0 0 0 1 0 0 0 0 0 1
patn3 T3 0 0 0 0 0 1 0 0 0 0 1
patn3 T4 0 0 0 0 0 0 0 0 0 0 1
grp effect 1 1 1 -1 -1 -1 0 0 0 0 0 X Ftest1: main effect group
T1-T4 1 0 0 1 0 0 0 0 0 0 0 X
T2-T4 0 1 0 0 1 0 0 0 0 0 0 X Ftest2: main effect time
T3-T4 0 0 1 0 0 1 0 0 0 0 0 X
C(T1-T4)-P(T1-T4) 1 0 0 -1 0 0 0 0 0 0 0 X
C(T2-T4)-P(T2-T4) 0 1 0 0 -1 0 0 0 0 0 0 X Ftest3: interaction group X time
C(T3-T4)-P(T3-T4) 0 0 1 0 0 -1 0 0 0 0 0 X
________________________________________
From: FSL - FMRIB's Software Library [[log in to unmask]] On Behalf Of Kyle Kern [[log in to unmask]]
Sent: Wednesday, March 18, 2009 8:01 PM
To: [log in to unmask]
Subject: [FSL] modeling slope over time
Hi,
I am trying to set up a longitudinal tbss analysis over 4 time points contrasting a patient
group to a control group. I'm interested in change over time rather than the effect of any
specifc time point. I was planning to model the slope for each subject and then put this
into a group t-test. Would modelling the slope for each subject be the same approach as a
simple correlation where my design matrix might look like this?:
-3
-1
1
3
Would this reflect the magnitude of the slope or does it only reflect the linearity like a
correlation coefficient? I would hypothesize that patients show a steeper slope than
controls though they may show more variability, so a least squares regression slope
might detect differences but a simple correlation coefficient might not.
Also, to perform a 2 level analysis with tbss is it preferable to use randomise and feed
the within subject t-stat into the group comparison or should I use the feat gui and treat
each within subject analysis as a timeseries and feed that into a higher level mixed
effects analysis?
thanks for all the help,
Kyle Kern
IMPORTANT WARNING: This email (and any attachments) is only intended for the use of the person or entity to which it is addressed, and may contain information that is privileged and confidential. You, the recipient, are obligated to maintain it in a safe, secure and confidential manner. Unauthorized redisclosure or failure to maintain confidentiality may subject you to federal and state penalties. If you are not the intended recipient, please immediately notify us by return email, and delete this message from your computer.
|