Subscribe to DSC Newsletter

Hi,

Is there a way to check for multi-collinearity of explanatory variables as part of model building using logistic regression?

I read somewhere that I should first use Linear Regression model to check for collinearity of explanatory variables and then remove / modify them before including them in logit model.

Are there any standard methods to handle this issue in logistic regression

Regards

Lakshmi

Views: 1261

Reply to This

Replies to This Discussion

Specialized methods to handle this issue in nonlinear regression like logistic regression do exist.  However, since collinearity is mainly an issue involving independent variables in a regression rather than the dependent variable or the link function between the independent and the dependent variables, using methods available for linear regression (for example, Belsley's collinearity diagnostics) is usually sufficient and applicable in nonlinear regression settings.

Thanks Matthew !

I was able to detect multicollinearity using tol and vif options in PROC REG and successfully resolved the issue !

Regards

Lakshmi

The COLLIN option of the MODEL statement in PROC REG includes condition indexes and variance proportions that may identify the set of independent variables involved in multicollinearity better than the TOL and the VIF options.  Try it out.  Sets of such variables with condition indexes exceeding 30 if the model has an intercept or 10 if it does not are the variables to concentrate on.

Thanks Matthew !

I will try it out

Regards

Lakshmi

RSS

On Data Science Central

© 2019   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service