help
help > RE: Questions about ANOVA models
Feb 18, 2015 04:02 PM | Alfonso Nieto-Castanon - Boston University
RE: Questions about ANOVA models
Hi Yong,
Sorry about this, the original contrast that I listed in my previous response for the OR-conjunction analyses was not correct, please use instead:
OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 -ones(1,10)/10; 0 -1 1 -ones(1,10)/10]
(I added an edit with this clarification to the original post response). If using each row of these contrasts separately, the results of those analyses should be similar to the simpler "Patients-pre' "Controls" (contrast -1 1) and "Patients-post" "Controls" (contrast -1 1) analyses, respectively (the only difference being in how the residuals of each analysis are being computed, which affects the statistics but not the effect sizes of your results; in the former case the residual variance is pooled across both patients-pre and patients-post samples, while in the latter analyses the residual variance is computed separately for the patients-pre or patients-post samples)
Hope this clarifies
Alfonso
Originally posted by Yong Li:
Sorry about this, the original contrast that I listed in my previous response for the OR-conjunction analyses was not correct, please use instead:
OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 -ones(1,10)/10; 0 -1 1 -ones(1,10)/10]
(I added an edit with this clarification to the original post response). If using each row of these contrasts separately, the results of those analyses should be similar to the simpler "Patients-pre' "Controls" (contrast -1 1) and "Patients-post" "Controls" (contrast -1 1) analyses, respectively (the only difference being in how the residuals of each analysis are being computed, which affects the statistics but not the effect sizes of your results; in the former case the residual variance is pooled across both patients-pre and patients-post samples, while in the latter analyses the residual variance is computed separately for the patients-pre or patients-post samples)
Hope this clarifies
Alfonso
Originally posted by Yong Li:
Hi
Alfonso,
I owe you a long overdue debt of gratitude. Your explanations are really helpful. Thank you!
Furthermore, I would like to better understand the 'conjunction of simple main group effects'. Is it possible to test 'differences between patients and control during pre- and post-medication respectively' with one manova design? Some preference to manova design is evident in RSfMRI literature. Does it make any sense to perform tests, e.g., 1. differences between patients and control during pre-medication by selecting "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 0 0 0 0 0 0 0 0 0 0]; 2. differences between patients and control during post-medication by selecting "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [0 -1 1 0 0 0 0 0 0 0 0 0 0]? If yes, what advantages and disadvantages of this design? After an attempt, I found different results between the aforementioned test 1&2 and those simple group effects by [1 -1].
I hope that I make myself clear. Thank you very much in advance as always.
Best regards,
Yong
Originally posted by Alfonso Nieto-Castanon:
I owe you a long overdue debt of gratitude. Your explanations are really helpful. Thank you!
Furthermore, I would like to better understand the 'conjunction of simple main group effects'. Is it possible to test 'differences between patients and control during pre- and post-medication respectively' with one manova design? Some preference to manova design is evident in RSfMRI literature. Does it make any sense to perform tests, e.g., 1. differences between patients and control during pre-medication by selecting "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 0 0 0 0 0 0 0 0 0 0]; 2. differences between patients and control during post-medication by selecting "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [0 -1 1 0 0 0 0 0 0 0 0 0 0]? If yes, what advantages and disadvantages of this design? After an attempt, I found different results between the aforementioned test 1&2 and those simple group effects by [1 -1].
I hope that I make myself clear. Thank you very much in advance as always.
Best regards,
Yong
Originally posted by Alfonso Nieto-Castanon:
...
Simple main group effects (e.g. comparing the patients and control groups connectivity measures after medication)
Select "Patients-Post" and "Controls" and enter contrast [1 -1]
Simple main group effects (e.g. comparing the patients and control groups connectivity measures before medication)
Select "Patients-Pre" and "Controls" and enter contrast [1 -1]
OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 0 0 0 0 0 0 0 0 0 0; 0 -1 1 0 0 0 0 0 0 0 0 0 0]
Note that the first test above (simple main medication effects) will be effectively equivalent to a paired t-test comparing pre- and post-medication connectivity measures, the next two tests will each be effectively equivalent to a two-sample t-test comparing patients vs. controls connectivity measures, and the last test will be equivalent to a manova of the between-group differences pre- and post- medication.
Simple main group effects (e.g. comparing the patients and control groups connectivity measures after medication)
Select "Patients-Post" and "Controls" and enter contrast [1 -1]
Simple main group effects (e.g. comparing the patients and control groups connectivity measures before medication)
Select "Patients-Pre" and "Controls" and enter contrast [1 -1]
OR conjunction of simple main group effects (differences between patients and control during either pre- or post-medication)
Select "Patients-Pre" "Patients-Post" "Controls" "Subject1" "Subject2" ... "Subject10" and enter contrast [-1 0 1 0 0 0 0 0 0 0 0 0 0; 0 -1 1 0 0 0 0 0 0 0 0 0 0]
Note that the first test above (simple main medication effects) will be effectively equivalent to a paired t-test comparing pre- and post-medication connectivity measures, the next two tests will each be effectively equivalent to a two-sample t-test comparing patients vs. controls connectivity measures, and the last test will be equivalent to a manova of the between-group differences pre- and post- medication.
...
Threaded View
| Title | Author | Date |
|---|---|---|
| Yong Li | Sep 15, 2014 | |
| Alfonso Nieto-Castanon | Sep 17, 2014 | |
| Yong Li | Feb 13, 2015 | |
| Alfonso Nieto-Castanon | Feb 18, 2015 | |
| Yong Li | Feb 24, 2015 | |
| Jeff Browndyke | Oct 1, 2014 | |
| Alfonso Nieto-Castanon | Oct 1, 2014 | |
| Traute Demirakca | Oct 2, 2014 | |
| Jeff Browndyke | Oct 1, 2014 | |