Dear SPM community,
I have some questions regarding the model I have employed to analyze a PET
memory retrieval study.
Study design:
3 Groups: Control, Schizophrenia 1, Schizophrenia 2
3 Conditions: Baseline, Recall 1, Recall 2
2 Runs per condition (3 x 2 runs in the entire experiment)
1 Covariate: Psychosis rating scale score (normal score = 0, maximum score
= 125)
Model:
- Multi-study, different conditions, with covariate
- Random effects model (creating straight means of the two runs per condition)
I have chosen:
- proportional scaling
- covariate: scores are not mean centered and not fitted
Analysis:
1) Within-group condition effect
0 1-1 0 0 0 0 0 0 0
0-1 1 0 0 0 0 0 0 0
(and other contrasts)
2) Main condition effect
0 1-1 0 1-1 0 1-1 0
(and other contrasts)
3) Main group effect
1 1 1 -1-1-1
1 1 1 -1-1-1
(and other contrasts)
4) Group * condition interaction
0 1-1 0-1 1 0 0 0 0
0 1-1 0 0 0 0-1 1 0
(and other contrasts)
5) Covariate effect
0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 -1
Questions:
1) Is the "straight mean" appropriate to create one condition file per
subject?
2) Is it appropriate not to mean center and not to fit the covariate?
3) Modeling the within-group main effect of condition in a 3 study design
(e.g., 0 1 -1 0 0 0 0 0 0 ) increases the degrees of freedom. Does that
mean that I should use this analysis rather then the single study design (0
1 -1)?
4) Analysis 3 works for me in SPM95 but not in SPM96 (I get the error
message: "Warning! Divide by zero.") How can I test for a main group effect
in a 3 study design in SPM96?
Your input is very much appreciated.
Stephan
-------------------
Stephan Heckers, MD
Department of Psychiatry, Massachusetts General Hospital
CNY-9132, Building 149, 13th Street
Charlestown, MA, 02129
Tel.: (617) 724 6141
Fax: (617) 726 4078
email: [log in to unmask]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|