2
$\begingroup$

I have 2 groups (tinnitus sufferers and controls) who are significantly different in age - I would normally control for age as a covariate (as it is a cognitive task) but it violates the assumption of ANCOVA in that the independent variable and covariate are related (as expected, older people are more likely to suffer tinnitus so this difference is expected).

I am worried that putting in age as a covariate removes my group effect because age and group are too closely related for them both to be in the model - can I legitimately not include this as a covariate based on this?

$\endgroup$

1 Answer 1

1
$\begingroup$

The standard way to handle this is to use an adjusted covariate: Instead of age itself, use age minus its group mean as the covariate. That way, each group has the same mean (namely zero) of the adjusted covariate.

In the analysis, the covariate-adjusted means will be the same as the unadjusted ones, but their standard errors will be smaller than if you had omitted the covariate, because you accounted for variations in the covariate.

$\endgroup$
1
  • $\begingroup$ You have made my day! This is just what I was after! :) $\endgroup$ Dec 4, 2014 at 0:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.