Dear all,

There is a question about ANCOVA that I’m not sure if it’s correct.

I want to compare the score of MMSE (a cognitive test) as IV between 3 groups.

Let’s say Normal, MCI (mild cognitive impairment), and AD (Alzheimer' s disease).

BUT according to the comparison of baseline, Age also has a significant difference between 3 groups. In order to reduce the influence of Age, I decided to put it as a covariate and control the comparison.

HOWEVER, when I read Field’s Discovering Statistics (2013) on pp. 400. It mentions that if the covariate has NO significant difference between IV, then the covariate will be appropriate to be used.

In my case, Age have significant difference between groups instead.

1. Does that mean I can NOT put Age as a covariate?

2. What is the name of this assumption? seems different from ‘homogeneity of regression slopes.

3. If I still want to control the influence of Age, which stats should I use?

Thank you in advance for any answers

Peter