Hi,

We have been discussing this topic on my lab for a long time. Is there any justification for reporting statistically significant results of several one-way ANOVAs when a previous and more adequate two-ways ANOVA had not reached significance?

Let's imagine: two conditions (i.e. rat strain and exposure to stress) that might impact one specific DV. The 2x2 analysis fails to shows either main effects or interaction (or shows main effects without interaction, or shows interaction without main effects), but an effect is suspected. Is it statistically justified to run one-way ANOVAs for each variable blocking the other (i.e.: rat strain 1 vs rat strain 2 under the same stress condition)?

We agree that a significant output of the whole analysis would allow several one-way's if you have some reason to avoid post-hocs, but this "negative" result gets us divided.

Thank you in advance!