Hi Berel,
If I have understood what Anderson wants to tell you is that if you can,
reduce the number of co-variates, if this is possible. Think that for every co-variate you enter in your model, you loos a Degree of Freedom and your sample is little to apply so many co-variates. For example, if you have 5 different vascular risk factors (Hypertension, Diabetes, Dyslipemia and Alcohol Consumption...) and they are highly correlated between them, something that in this case, use to happen, then you should try to reduce them into one or two variables only...and so on for the other co-variates...In order to know this, you can run bivariate correlations and see how your data behaviour regarding their relations/correlations between them...and between your variables of interest. Once you have seen this, you can try to add those variables into a PCA model and just extract the components whose weight is higher than one...this will help you to be more "parsimoniuos" taking into account the size of your data.
You can find some good explanations about how to do this with SPSS in Google...you can extract Factors and save them as the new variables you have chosen for covariation. Also, as Anderson recommended me, be aware of circularity in your post-hoc analyses....
Hope this helps...
