1

I realize that you can both go forward or backward, or even in both directions, however I'm finding it a little confusing when one is more appropriate than the other? Can somebody explain to me or give me examples when it's appropriate to use one or the other or both?

WoeIs
  • 121
  • 1
  • 1
    If you plan on doing inference after stepwise regression, any method [will give invalid results](https://stats.stackexchange.com/q/179941/1352). if you want to do prediction, I do not know of a general rule. Except to rather try regularization, say via the lasso. – Stephan Kolassa Jun 06 '18 at 15:57
  • 1
    As @StephanKolassa alluded, stepwise regression is unsafe and outdated. Like the [penny-farthing](https://en.wikipedia.org/wiki/Penny-farthing), it is has been replaced by better, more modern techniques. I wouldn't do stepwise regression unless some class is bizarrely forcing you to. – Matthew Gunn Jun 06 '18 at 16:55

0 Answers0