On Monday, April 08, 2013 12:41:55 you wrote:
> Hi all,
>
> we have DTI data (axial diffusivity) comparing three independent groups
> that underwent three different treatments. We did an F test and contrasts
> between pairs for groups. The contrasts indicate significant effects
> between subgroups, the F test only showed a trend result (was not
> significant). Does anyone have a suggestion how one can justify to look at
> contrasts between individual groups even if the overall F test did not come
> out significant?
If you have an a priori reason to be interested in some contrasts more than others, then you might be able to justify reporting the contrast of interest even if the F-test was not significant. For example, suppose I collect the height, number of eyes, number of digits, number of meals eaten per day, etc. of a small sample of men and women. An F-test might not reveal any significant difference between men and women. However, even before collecting my data, I only expected to see a difference in the height variable. It would therefore be appropriate to report just the significant difference in height between men and women.
> I feel that the results from the contrasts should be
> reported because they are meaningful.
You should be fine so long as you report all of the results (not just the significant ones) and state that the F-test was not significant.
|