Hi all, newbie here with a question I was hoping someone with more statistical theory knowledge could answer.

I have a question for you re: the use of post-hoc tests on ANOVA data.

My manuscript is back to me for a second round of revisions and I am a bit confused by a reviewer request. This is a retrospective study where we compared 7 different people (physicians) who performed a procedure. We looked at different metrics of performance/clinical outcomes at specific time points after the procedure was performed. I performed an ANOVA on both the main effect (time) and between physicians.

As a group, ANOVA data suggest there were no significant differences between physicians but the reviewer asked for post-hoc analysis (in a more informal request) of this ANOVA data. When I ran the post-hoc analysis, there were cases among the 5-6 metrics observed (pain, activity, etc.) where there was 1 significant (p < 0.05) difference between 2 of these physicians despite a global ANOVA showing a p = 0.17 or so.

Questions:
1) Is a post-hoc analysis valid when the global ANOVA was not significant?
2) Is it possible to have scattered significant post-hoc interactions when the global ANOVA comparing all 7 of these physicians is not significant? I am a bit confused by this...

Thanks for your help!

Bob