What does theory say concerning the relationship between anxiety and grade point average? Is it typically found to be linear? Curvilinear? To my knowledge, GPA is generally considered an interval scale, so ordinal regression wouldn't be appropriate. There's no rule against conducting an analysis with both linear and nonlinear (polynomial) models and seeing which describes the data best. *Just be careful about overfitting.

Hi Frieder, as a 6-12 former teacher and as a psychometrist, and statistician, the way you operationalize your outcome (GPA) will tell you the type of statistical methods you could use. If looking at letter grades (e.g., A, B, C), then, yes, your outcome is ordinal. But if you look at GPA expressed with numbers (rational numbers; e.g., 3.5) using a 4.0 scale, then, your outcome is an interval scale (i.e., there is the same distance from 2.0 to 3.0 and from 3.0 to 4.0), as Dr. Tomkings also explains. I hope this is helpful.

Dear Frieder L Schillinger ,

Following on from Patricia Rodriguez de Gil and Blaine Tomkins 's points, just because you can treat your dependent variable as scale doesn't mean you should use a linear regression model. You should examine the dispersion of your DV against the residuals of each IV in a linear regression model and decide whether they are normally distributed. Alternatively, create a scatter plot of each IV against your DV in turn and look for a "cigar" shape. Please see my study guide on simple linear regression.

What if I want to categorize my GPA into two groups?

For example: students with a higher GPA and students with a lower GPA.

Is GPA still considered ordinal data or is it better expressed as nominal data?