The best way to check for validity is to collect new data. If the old model still fits the new data, then the regression model is a good fit and valid.


Different methods for checking validity against new data:

1. Re-estimating the model using the new data. The old and new regression models are compared for consistency. Check the regression coefficients, coefficient of determination, normality plots, and residual plots. If the results of both are consistent then the new regression model is strong.
2. Conclude from the new data all of the "good" models and obtain the best model. If the model originally selected with the old data is the same as the new one, then the model is efficient under the new conditions.
3. Measure the actual predictability by using the model to predict each case in the new data set and calculate the MSPR (mean squared prediction error) using this formula:

Compare the MSE to the MSPR. 

• If the MSPR is close to the MSE, then the MSE gives a good predictability of the model. 
• If the MSPR is much larger than the MSE, then rely on the MSPR to indicate how well the selected regression model will predict in the
  future.

 Return to Building a Regression Model

Regression Tutorial Menu    Dictionary

STATS @ MTSU