Variables suitable for Linear Regression?

Hey everyone!

I am trying to find out if two of my variables (dependent: continous / independent: categorical) are suitable for linear regression with STATA. Usually it is suggested to create a scatterplot with a regression line but as far as i know this is only possible if both of the variables are continous. Is there any other easy way to find out about linearity?

Thanks in advance!!


Active Member
If the independent variable is categorical we are imposing a linear relationship with the dependent variable. In this way, we can investigate the mean change in the dependent variable relative to the baseline level in the categorical predictor. You might find useful information searching for ANOVA. It is an alternate way to formulate the OLS regression model. Boxplots and histograms would be helpful visualizations.


No cake for spunky
It might be noted researchers do not agree on this issue. Economists have decided for example that categorical dependent variables are commonly fine in linear regression in most cases although other disciplines have conniption fits about that. They call these linear probability models.

If you are using them as predictors they are fine period, although it is best (according to many) to make these dummy variables. If they are being used as the DV you have to decide if the economist or other researchers are right (and that depends on your field). :p

I have no idea why you would generate a scatter plot. Are you testing for normality of the variables, or the need to transform them?

This has nothing to to with Stata. It would be true in any software.