help
help > RE: Exchange block and contrast questions
Jan 19, 2017 01:01 AM | Jeff Peters
RE: Exchange block and contrast questions
Hi Andrew,
Thank you so much for clearing up my confusion about the exchange block. That makes perfect sense now.
Regarding the contrast question, my apologies for being unclear. What I meant to say was that I have created the contrast for task condition A > task condition B, now I would like to create the contrast for task condition B > task condition A (as I don't think they will give the same results). I'm just unclear what the second contrast vector looks like (given that the first one is [2 1 0...]).
Jeff
Originally posted by Andrew Zalesky:
Thank you so much for clearing up my confusion about the exchange block. That makes perfect sense now.
Regarding the contrast question, my apologies for being unclear. What I meant to say was that I have created the contrast for task condition A > task condition B, now I would like to create the contrast for task condition B > task condition A (as I don't think they will give the same results). I'm just unclear what the second contrast vector looks like (given that the first one is [2 1 0...]).
Jeff
Originally posted by Andrew Zalesky:
Hi
Jeff,
The exchange block vector is simply used to identify the same subject at different time points in a repeated measures design. For example , suppose the rows of your design matrix are ordered such that: S1T1, S2T1, S1T2 S2T2, where S1TI means time point 1 (T1) of subject 1 (S1). In this case, the exchange blocks would be: [1 2 1 2], since the first subject is positioned at rows 1 and 3, and the second subject is positioned at rows 2 and 4.
Alternatively, if the rows were ordered such as S1T1, S1T2, S2T1 S2T2, then the exchange block would be [1 1 2 2].
I'm not sure if I follow your 2nd question about "reverse contrasts". Seems like a complicated contrast. The FSL mailing list has some discussion of the this example.
Andrew
Originally posted by Jeff Peters:
The exchange block vector is simply used to identify the same subject at different time points in a repeated measures design. For example , suppose the rows of your design matrix are ordered such that: S1T1, S2T1, S1T2 S2T2, where S1TI means time point 1 (T1) of subject 1 (S1). In this case, the exchange blocks would be: [1 2 1 2], since the first subject is positioned at rows 1 and 3, and the second subject is positioned at rows 2 and 4.
Alternatively, if the rows were ordered such as S1T1, S1T2, S2T1 S2T2, then the exchange block would be [1 1 2 2].
I'm not sure if I follow your 2nd question about "reverse contrasts". Seems like a complicated contrast. The FSL mailing list has some discussion of the this example.
Andrew
Originally posted by Jeff Peters:
Dear Andrew,
I have a within-subject design with 21 subjects and 3 task conditions A, B, and C. Using the FSLwiki (triple t-test section), I was able to create the design matrix (would you mind taking a quick look to see whether it looks fine). Before running the analysis, I have a couple of questions to which I would be grateful for answers:
Contrast: I have the following contrasts: A-B=[2 1 0...], B-C=[-1 1 0...], A-C=[1 2 0...]. I am also interested in the reversed contrasts (i.e., B-A, C-B, and C-A). Going by what I think is the logics behind contrast setup, I came up with: B-A=[-1 -2 0...], C-B=[-1 1 0...], C-A=[-1 -2 0...]. The C-B contrast seems incorrect as it is the same as B-C. Would you kindly point out where I may have gone wrong here?
Exchange block: from the manual, I see that the exchange block is required for repeated measures within-subject designs, which is what I have here. However, I am confused as to what input goes into the vector. Since I have 21 subjects, the vector will have 21 rows (though a search throughout the forum doesn't always show that users have stuck to this principle). Am I missing something? To really address my ignorance, could you please explain the reason for constraining the permutations?
Best,
Jeff
I have a within-subject design with 21 subjects and 3 task conditions A, B, and C. Using the FSLwiki (triple t-test section), I was able to create the design matrix (would you mind taking a quick look to see whether it looks fine). Before running the analysis, I have a couple of questions to which I would be grateful for answers:
Contrast: I have the following contrasts: A-B=[2 1 0...], B-C=[-1 1 0...], A-C=[1 2 0...]. I am also interested in the reversed contrasts (i.e., B-A, C-B, and C-A). Going by what I think is the logics behind contrast setup, I came up with: B-A=[-1 -2 0...], C-B=[-1 1 0...], C-A=[-1 -2 0...]. The C-B contrast seems incorrect as it is the same as B-C. Would you kindly point out where I may have gone wrong here?
Exchange block: from the manual, I see that the exchange block is required for repeated measures within-subject designs, which is what I have here. However, I am confused as to what input goes into the vector. Since I have 21 subjects, the vector will have 21 rows (though a search throughout the forum doesn't always show that users have stuck to this principle). Am I missing something? To really address my ignorance, could you please explain the reason for constraining the permutations?
Best,
Jeff
Threaded View
| Title | Author | Date |
|---|---|---|
| Jeff Peters | Jan 18, 2017 | |
| Andrew Zalesky | Jan 19, 2017 | |
| Jeff Peters | Jan 19, 2017 | |