I have made a generalised linear model with a single response variable (continuous/normally distributed) and 4 explanatory variables (3 of which are factors and the fourth is an integer). I have used a Gaussian error distribution with an identity link function.

Do I need to check for multicollinearity and interactions amongst explanatory variables? If yes, how do I do this with categorical explanatory variables?

  • 2
    $\begingroup$ If you plan on interpreting the individual $p$-values, you should check for collinearity. Checking for collinearity is a frequently discussed topic on this website so doing a search for 'collinearity' will probably be fruitful. To look at dependence between categorical variables you can look at the usual $\chi^2$ tests or something like that. In this thread,stats.stackexchange.com/questions/8088/…, the problem is discussed for binary predictors $\endgroup$
    – Macro
    Commented Jul 15, 2012 at 13:23

1 Answer 1


If the link is the identity function this model is just ordinary regression. But even for other link functions multicollinearity if it is a serious problem then you might consider subset selection. If variables are suspected to interact consider including interaction terms in the model. There is no problem including categorical variables in such models.

  • $\begingroup$ many thanks for superb advice - well impressed with this site in short time I have been using it. $\endgroup$
    – luciano
    Commented Jul 15, 2012 at 18:46
  • $\begingroup$ Check out the material on hierarchical regression and causality at stats.stackexchange.com/questions/10020/… $\endgroup$
    – rolando2
    Commented Jul 16, 2012 at 5:54

Your Answer

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

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