Use backward elimination procedure to decide which predictor variables can be dropped from the regression model. Control the type I error at = . 10 at each stage

In using backward elimination procedure how to control for type I error?

In backward elimination, one variable gets dropped at each stage. This could lead to a type 1 error for a given α (say 0.05). How to calculate this error at each step? Please put details in the explanation.

I know how to calculate type I error in like a t-test setting, but I do not get how to apply the same concept to backward elimination.

  • $\begingroup$ "Control the type I error" is a (misleading) turn of phrase; it doesn't literally mean that you are trying to control errors in the procedure overall. You know that backwards elimination uses a nominal p-value threshold to determine which variables to remove at each stage. This phrase merely specifies that threshold. $\endgroup$ – whuber Mar 9 '18 at 15:23
  • $\begingroup$ One of the several problems with backward elimination (and forward and stepwise) is that type I error is inflated and by unknown amounts. There's no good way to fix this. However, this question appears to ask you to control it at each stage. You do that by seting an option in the software. $\endgroup$ – Peter Flom Mar 9 '18 at 15:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.