Hi Anderson,
Thank you very much for the feedback. I followed your orientations and I removed the unnecessary contrasts.
Regarding the F-test. When I choose the check box beside the contrast the get the attached error dialog box. Kindly what is the meaning of this message and what I am doing wrong?
(I attached the error message)
Jon
Hi Jon,
They are not wrong, only could be a lot simplified. The design itself is fine, but note the large correlations between EVs of interest and nuisance EVs, that will invariably lead to loss of power.
The contrasts can be changed:
C1-C12 are fine and can be left as is. However, note that the labels don't match what these are testing. E.g., C7-C8 are testing A vs. D, but their labels incorrectly say A vs. C. The opposite happens with C9-C10.
C13-C16 are not needed. These will test if FA > 0, but you know this already. Same for F1 that combines all these.
C17-C28 are repetitions of the earlier ones and can all be dropped.
F2-F5 are all testing the same thing, will produce the exact same test statistic and, if enough permutations are done, will lead to the exact same p-values, and all can be removed and replaced for a single new F-test that uses your current C1, C7 and C9 (ignoring the incorrect labels).
All the best,
Anderson
On 5 September 2016 at 15:26, John anderson <[log in to unmask]> wrote:
Hi Anderson,
I would like to inquire if the attached model is correct.
I have four groups (A, B, C and D).
The subjects scanned MR/DTI. I want to study the difference between the groups in fractional anisotropy using TBSS.
I built the attached model, and I included (age, gender and weight) as a covariates in this model. In order to estimate F statistics. I created contrast for each condition.
Kindly is this model correct?
Do you kindly suggest me any additional steps to improve it or fix it?
Your help and support is highly appreciated!
Jon
|