how does fitglm treat categorical variables?

Dear all,

I'd like to verify something about categorical variables in fitglm. As input I use a table which also contains categorical variables (sex (0 vs 1), and education (on a scale 1 to 7). If I understand the documentation correctly, fitglm automatically treats these as categorical, and it also automatically dummy codes when necessary? (for education in this case). Is this correct?

Thanks very much.

Best

Hans van der Horn

Answer 1

[Jeff Miller]

I don't that is correct. WIth numerical values in the table, I think you have to mark the variables as categorical with something like:

tbl.sex = categorical(tbl.sex);
tbl.education = categorical(tbl.education);

Where there are three or more categories (e.g., education) you can check the df's to make sure that the variable has been treated as categorical (6 dfs) rather than numerical (1 df, essentially a regression slope across 1-7).

[Hans van der Horn]

Dear Jeff,

Thanks for your answer. I tried as you suggested and indeed it works better, with now dummy variables defined in the model.

Best Hans

[Jeff Miller]

Hi Hans,

That sounds good. If this answer solves the problem that you were having, then please accept it using the "Accept" button (so that the question no longer appears to be open).

Jeff