# Relevance of covariates?

#### hebertodelrio

##### New Member
I have a big group $$A$$ of individuals that have one common characteristic that make them part of a study in which certain binary variable has to be accounted for. Logistic regression can be used in this case. This group of individuals can be separated in disjoint subgroups $$A_n$$ based on the date they entered the study. When I perform logistic regression certain covariates appear to be statistically significant to explain log odds ration, but when I do perform the same analysis in each of the subgroups $$A_n$$ most of the covariates that were significant globally are no longer significant, is there a good explanation for this?

Last edited:

#### staassis

##### Member
Yes, those covariates are correlated with when the subject entered the study.

#### hlsmith

##### Omega Contributor
OR, when you run the sub-analyses the sample sizes are smaller and you no longer have the power to discern differences. Present the model outputs to better help us understand the results.