help > NBS ANCOVA 3 Group Contrasts
Showing 1-2 of 2 posts
Jan 5, 2021 10:01 PM | Jessica Hua
NBS ANCOVA 3 Group Contrasts
Hi NBS Experts,
I had a question regarding contrasts in my study design. My study includes three groups: HC, PR, SCZ and the covariate is age. My abbreviated design matrix is:
1 1 0 21
1 1 0 22
1 1 0 24
1 1 0 20
...
1 0 1 17
1 0 1 16
1 0 1 18
1 0 1 21
...
1 0 0 22
1 0 0 22
1 0 0 18
1 0 0 24
...
To examine an overall difference among groups would my contrast be [0 1 1 0]? And how would I write the contrasts to compare HC > SCZ, HC > PR, PR > SCZ?
Thank you,
Jessica
I had a question regarding contrasts in my study design. My study includes three groups: HC, PR, SCZ and the covariate is age. My abbreviated design matrix is:
1 1 0 21
1 1 0 22
1 1 0 24
1 1 0 20
...
1 0 1 17
1 0 1 16
1 0 1 18
1 0 1 21
...
1 0 0 22
1 0 0 22
1 0 0 18
1 0 0 24
...
To examine an overall difference among groups would my contrast be [0 1 1 0]? And how would I write the contrasts to compare HC > SCZ, HC > PR, PR > SCZ?
Thank you,
Jessica
Jan 6, 2021 11:01 AM | Andrew Zalesky
RE: NBS ANCOVA 3 Group Contrasts
Hi Jessica,
To test for a difference between any of the three groups (i.e. three-way ANOVA), the first column of the design matrix should only include 1's for the first group (rather than all groups). The contrast would then be [1 1 1 0]. Select F-test.
That said, I think that your current design and the contrast of [0 1 1 0] would give the same result. You can check this for yourself.
Three independent t-tests can be used to asses specific pairs of groups. The NBS Manual and Help section provides examples of a t-test.
Another option is to perform post-hoc t-test just on the connections that were identified with the initial F-test. This is a bit more complicated and would require you to extract the connectivity values from each edge comprising the significant network.
I hope that helps and I would be glad to answer any further questions!
Best wishes,
Andrew
Originally posted by Jessica Hua:
To test for a difference between any of the three groups (i.e. three-way ANOVA), the first column of the design matrix should only include 1's for the first group (rather than all groups). The contrast would then be [1 1 1 0]. Select F-test.
That said, I think that your current design and the contrast of [0 1 1 0] would give the same result. You can check this for yourself.
Three independent t-tests can be used to asses specific pairs of groups. The NBS Manual and Help section provides examples of a t-test.
Another option is to perform post-hoc t-test just on the connections that were identified with the initial F-test. This is a bit more complicated and would require you to extract the connectivity values from each edge comprising the significant network.
I hope that helps and I would be glad to answer any further questions!
Best wishes,
Andrew
Originally posted by Jessica Hua:
Hi NBS Experts,
I had a question regarding contrasts in my study design. My study includes three groups: HC, PR, SCZ and the covariate is age. My abbreviated design matrix is:
1 1 0 21
1 1 0 22
1 1 0 24
1 1 0 20
...
1 0 1 17
1 0 1 16
1 0 1 18
1 0 1 21
...
1 0 0 22
1 0 0 22
1 0 0 18
1 0 0 24
...
To examine an overall difference among groups would my contrast be [0 1 1 0]? And how would I write the contrasts to compare HC > SCZ, HC > PR, PR > SCZ?
Thank you,
Jessica
I had a question regarding contrasts in my study design. My study includes three groups: HC, PR, SCZ and the covariate is age. My abbreviated design matrix is:
1 1 0 21
1 1 0 22
1 1 0 24
1 1 0 20
...
1 0 1 17
1 0 1 16
1 0 1 18
1 0 1 21
...
1 0 0 22
1 0 0 22
1 0 0 18
1 0 0 24
...
To examine an overall difference among groups would my contrast be [0 1 1 0]? And how would I write the contrasts to compare HC > SCZ, HC > PR, PR > SCZ?
Thank you,
Jessica