Analytics/Statistics

tcj9
StepstoEvaluateaRegressionModel.docx
Home>Mathematics homework help>Statistics homework help>Analytics/Statistics

Steps to Evaluate a Regression Model

Good/Bad

1. F-test

2. T-test for each independent variable

3. Look at the residual plot—should be random; (if time series data look at D-W test (want to not reject; otherwise significant first order autocorrelation))

4. Multicollinearity (VIF);

1≤ VIF≤ 5 no significant multicollinearity

5 < VIF ≤ 10 be concern that some multicollinearity may exist

VIF > 10 significant multicollinearity

How good of a model

1. Adjusted R2—most important when looking at relationships

2. RMSE (or standard error) –most important when forecasting