Hi, I have a question regarding ANCOVA. I feel it might be a stupid one but there you go. I have 2 groups (men and women) for which I am scoring a behaviour. I also have the age of the subjects but it so happens that women are older than men to the extent that there is almost no overlap between the 2 age groups.
I know that to run an ANCOVA, the covariable has to be independent from the explanatory factor. In this case, it is not true clearly as there is a significant relationship between gender and age (female older than male).
What worried me more before analysing anything of course was the fact that if the age is different between the groups what is the point of looking at the relationship age/scores in the first place since the age range is different in each gender?
But is it not the same thing at the end of the day? I mean if the covariate is not independent from the factor, by essence it means that the range of the covariable is going to be different within each group.
It sounds a bit metaphysical now that I have written it but it is Friday afternoon so ...
Anyway, thanks to anyone who can answer me!


TS Contributor
As a self proclaimed metaphsycist, heres my take,

I am not sure it is strictly necessary for the covariate(age) to be independant from expanatory variable(gender). At any rate, you could surely plug the data into your computer and get a p value for any set of data. The difficulty, as you have identified is in interpretting the p-value from such an analysis.

If you were to run a one way anova without covariate and observe an effect, and then run the ancova and observe no effect, it would suggest that age was a common cause of both your outcome and gender. It is queer to think of age as a cause of gender, so maybe just say they are associated somehow? In epidemiology this is called confounding, sometimes called "simpsons paradox".

At any rate, I think your skepticism is warranted. It would be odd to say that gender had an effect within the strata(after adjusting for) age when the men and women have no common ages.