09-06-2014 02:22 PM
Regress price, ad dollars, ad frequency, other brand prices on ln(sales) using PROC GLM. In this case, add brand(dummy variable) to the regression. Save regression parameters and model predictions of ln(sales) (with appropriate confidence intervals) to two comma delimited files. Generate in-sample and validation sample fit statistics: r-square, RMSE and save to external file.
For the predictions, create one file for each predictor (excluding brand) where the predictions cover the range of observed values. Hold all other predictors at their mean value.
To generate the predictions of sales, I use "ODS output predictedvalues= file"; and also add clm and p options behind the proc glm statement. I only got all the estimated observed values but not for each predictors. I was wondering how to do that?