The Right Regression

gpreece — Wed, 08 Nov 2017 21:38:53 GMT

Hello SAS Community!

I need to hear your thoughts regarding the best regression choice for my model.

Sample size is 15K

I have a dependent variable comprised of 12 items - all yes or no. The score can be between 0-12 all assigned equal values.

Further, my model consists of 17 predictor/independent variables. Variable range from gender (binary), income (categorical), to risk aversion scale (interval).

I initially planned to run an OLS Regression but I am getting conflicting opinions on whether or not that is the right option. Your opinions and advice is greatly appreciated.

Thanks,

Gloria

Re: The Right Regression

PaigeMiller — Wed, 08 Nov 2017 23:46:55 GMT

You will have problems with OLS since your independent variables are almost definitely correlated with one another.

Partial Least Squares regression will provide better model fits (lower mean square error of the coefficients, lower means square error of the predicted values) than OLS in this case, and will handle the collinearity between the predictors better than OLS will handle it.

topic Re: The Right Regression in Statistical Procedures

The Right Regression

Re: The Right Regression