# Thread: Binomial logistic regression linearity assumption with categorical variables ?

1. ## Binomial logistic regression linearity assumption with categorical variables ?

I am trying to run a binomial logistic regression, and would like to check the linearity assumption. However all the resources I'm reading discuss the procedure for continuous variables, and I am only working with ordinal/categorical predictors. Can someone take me through it step by step, or lead me to a resource that explains the procedure ?

2. ## Re: Binomial logistic regression linearity assumption with categorical variables ?

Are there k-1 dummy variables being created to code your ordinal/categorical variables?

3. ## Re: Binomial logistic regression linearity assumption with categorical variables ?

In linear regression the assumption of linearity with categorical predictors is always met so you don't have to worry about it. Is this true with logistic regression (where the assumption deals with the predictors and the logit or odds ratio not the raw data I believe to start with)?

What is a good test of linearity with a continuous predictor in logistic regression?

4. ## Re: Binomial logistic regression linearity assumption with categorical variables ?

The predictor variables all have been coded in a similar way from a self-report scale where 1=never 2=sometimes 3=often, etc.

5. ## Re: Binomial logistic regression linearity assumption with categorical variables ?

I have read that you are supposed to conduct a Box-Tidwell test for continuous variables, but it does not seem to apply to ordinal/categorical variables.

6. ## Re: Binomial logistic regression linearity assumption with categorical variables ?

Originally Posted by Maggierose
The predictor variables all have been coded in a similar way from a self-report scale where 1=never 2=sometimes 3=often, etc.
You need to code the variables as dummies, this will take care of the problem. The dummies always have two levels only.
regards

 Tweet