Hi Andrew,

I'm new to NBS and I would like to do a 2x2 mix effect ANOVA analysis. I have two groups of (control and patient) and they were measured at two different time point (A and B). I want to evaluate that there is no effect of time such that the connectivity would be relative stable  but there is an effect of group difference which connectivity differ between group

Sincerely,

Thomas

Hi Thomas,

please refer to one of the previous posts on this board. There are many posts that describe the design matrix and contrast for a 2 x 2 ANOVA with one repeated measure (time in this case).

Let me know if you have any specific questions,

Andrew


Originally posted by Thomas Tan:

Originally posted by Andrew Zalesky: Hi Andrew,

I've dig into some of the previous posts and I came up with design matrix for my study. For example, 2 groups 2 subjects per group and across two time points. PT denote for patient and CT denote for control, T1 = Time 1 and T2 = Time 2.

CT1T1 0 1 1 0 0 0
CT1T2 1 0 1 0 0 0
CT2T1 0 1 0 1 0 0
CT2T2 1 0 0 1 0 0
PT1T1 0 0 0 0 1 0
PT1T2 1 1 0 0 1 0
PT2T1 0 0 0 0 0 1
PT2T2 1 1 0 0 0 1

1st col is main effect of time
2nd column is interaction between group and time
last 4 for mean of each subject
Contrast (for interaction)
0 1 0 0 0 0
or: 0 -1 0 0 0 0

Exchange blocks
1 1 2 2 3 3 4 4

If I do this, would it mean that I'm looking at the group x time interaction. If there is no interaction, I can perform follow-up to test the main effect of group (PT > CT at time 1  and PT > CT at time 2) separately. In addition, if I use the contrast for the main effect of time is either: [1 0 0 0 0 0] or [-1 0 0 0 0 0] with the exchange blocks above , does it mean that it will determine whether the control/patient group would differ across two time point?

Hi Thomas,

If you are only interested in the group difference (patient vs control), you don't need two different time points. I suggest simply averaging the data across the two time point for each subject and then running a t-test.

An alternative approach is to remove the first and second column in the design matrix that you have quoted below and then add a column with 1's and 0's which encodes patients versus controls. You would still keep the last 4 columns corresponding to the means for each subject.

CT1T1 1 1 0 0 0
CT1T2 1 1 0 0 0
CT2T1 1 0 1 0 0
CT2T2 1 0 1 0 0
PT1T1 0 0 0 1 0
PT1T2 0 0 0 1 0
PT2T1 0 0 0 0 1
PT2T2 0 0 0 0 1

Contrast: [1 0 0 0 0 ] or [-1 0 0 0 0]

This would test for a between-group difference in patients and controls.

Andrew






Originally posted by Thomas Tan:

Thank you,

If my hypothesis 1 is that the connectivity would be the same in control group (CT1 = CT2), and my hypothesis 2 is that (PT1 = PT2), would i perform t-test separately in each cases?  Can you elaborate on this

1. Hypothesis 1 ( CT1T1 = C1T2 )

CT1T1 1 1 0 0 0
CT1T2 0 1 0 0 0
CT2T1 1 0 1 0 0
CT2T2 0 0 1 0 0

contrast [1,0,0,0] or [-1,0,0,0]

Hi Thomas,

I'm not sure what you mean by "CT1T1".

If you are simply interested in testing for a difference between the two time points only in the control group, then the patient data can be omitted and you can do a t-test (repeated measures) only using the control data.

You could then perform the same test for the patient group.

Andrew

Originally posted by Thomas Tan:

Hi Andrew,

Sorry for the confusion when i said CT1T1, i meant control from phase 1, and T2 is phase 2
Therefore, if i am interested only in testing for a difference between the two time points only in the control group, my DM, contrast and exchange block should be looking like this?

Hi Thomas,

I'm not sure what you mean by "CT1T1".

If you are simply interested in testing for a difference between the two time points only in the control group, then the patient data can be omitted and you can do a t-test (repeated measures) only using the control data.

You could then perform the same test for the patient group.

Andrew

Originally posted by Thomas Tan: [/quote]

Hi Thomas,

A within-subject t-test would be sufficient, if you are only interested in the time difference between the two time points in the control group.

The design matrix for three subjects would be:

C1T1  0  1  0  0
C1T2  1  1  0  0
C2T1  0  0  1  0
C2T2  1  0  1  0
C3T1  0  0  0  1
C3T3  1  0  0  1

The contrast would be:
[1 0 0 0] or [-1 0 0 0]

Select: t-test

The exchange blocks would be:
[1 1 2 2 3 3]

Andrew







Originally posted by Thomas Tan:

Thank you very much,

Thomas