Dear experts,
I am so sorry for the long e-mail and I apologize in advance for the repetitive and simple questions, but after doing the tutorial and reading the manual, I am still not sure if I set the parameters right in SwE.
I have a configuration similar to Brunno's one: 2 groups (15 subjects each) and 3 time points. And I would like to answer the following questions:
1) Is there any effect along with time within each group (within-subjects)?
I set the following:
For group1 (G1):
Scans: 45 files (G1T1(15) G1T2(15) G1T3(15))
Groups: ones(1,45)
Visits: kron(1:3,ones(1,15))
Small sample adjustments: type 3 (should I use C2?)
Degrees of freedom: approx III
Subjects: kron(ones(1,3),1:15)
I did the covariates as suggested:
Covariate 1
vector: kron([1 0 0 0 0 0],ones(1,15))
name: G1T1
Covariate 2
vector: kron([0 1 0 0 0 0],ones(1,15))
name: G1T2
Covariate 3
vector: kron([0 0 1 0 0 0],ones(1,15))
name: G1T3
a) For group2 I did the exact same thing but using the scans from group2, because I think that adding all the 6 covariates (3 for G1 and 3 for G2) at this point will affect the model in the wrong way (by reducing the degrees of freedom), won't it? So, is the setting of parameters correct?
b) Does SwE by itself includes the covariates to remove the individual mean effects? I mean, does it somehow remove the effect of subject 1 alone (the average of his 3 visits), subject 2 alone (the average of his 3 visits), subject 3, ..., subject 15? Or should I include covariates like the ones below:
Covariate 4:
vector: kron(ones(1,3),[1 0 0 0 0 0 0 0 0 0 0 0 0 0 0])
name: Subj1
Covariate 5:
vector: kron(ones(1,3),[0 1 0 0 0 0 0 0 0 0 0 0 0 0 0])
name: Subj2
...
Covariate 18:
vector: kron(ones(1,3),[0 0 0 0 0 0 0 0 0 0 0 0 0 0 1])
name: Subj15
c) Than my contrasts will simply be: [1 -1 0], [-1 1 0], [0 1 -1], [0 -1 1], [1 0 -1], [-1 0 1], right? Do I need to correct for multiple comparisons?
2) Is there any difference between groups along with time (within-subject + between group comparisons)?
I set the following:
Scans: 90 files (G1T1(15) G1T2(15) G1T3(15) G2T1(15) G2T2(15) G2T3(15))
Groups: kron(1:2,ones(1,45))
Visits: [kron(1:3,ones(1,15)) kron(1:3,ones(1,15))] (Is this right? Because Brunno's one was different...)
Small sample adjustments: type 3
Degrees of freedom: approx III
Subjects: [kron(ones(1,3),1:15) kron(ones(1,3),16:30)]
Covariate 1
vector: kron([1 -1 0 0 0 0],ones(1,15))
name: G1T1>T2
Covariate 2
vector: kron([0 1 0 0 0 0],ones(1,15))
name: G1T1<T2
...
Covariate 6
vector: kron([0 0 1 0 0 0],ones(1,15))
name: G1T1>T3
Covariate 7
vector: kron([0 0 0 1 0 0],ones(1,15))
name: G2T1>T2
...
Covariate 12
vector: kron([0 0 0 0 1 0],ones(1,15))
name: G2T3>T1
Than my contrasts will be:
Contrast1: if the increase from T1 to T2 in G1 was greater than in G2.
vector:[1 0 0 0 0 0 -1 0 0 0 0 0]
Contrast...
a) Is this correct?
3) If I want to check correlation between the change from T1 and T2 in group 1 with their score in some scale, how would I set the parameters and the constrast in SwE?
I do apologize for the large message, and I really appreciate any help.
All the best,
Raphael
|