Significant Main Effect in Omnibus, But Not In Post Hoc

Greetings,

I'm analyzing my dependent variables using a 3 (Condition) x 4
(Ethnicity: Caucasian vs. Asian vs. African-American vs. Other) x 2
(Gender: Male vs. Female) multivariate analysis of covariance. The
data reveal a significant main effect for ethnicity that does not show
up in post hoc analyses using a Bonferroni adjustment. I've been using
the Bonferroni adjustment for all my analyses (except for my condition
analyses for which I am using the LSD correction on the recommendation
of Levin, Serlin and Seaman, Psych Bull., 1994) and want to continue
using it for this analysis for the sake of consistency. My question
is, do I have a meaningful main effect that needs to be reported in my
write up even if no pairwise comparisons are significant and the
ethnicity variables are not central to my hypothsis?

sb

.

Relevant Pages
Loading