0

To select a model I use stepwise with the likelihood ratio test but I have seen that people use the z test (the one that appears in R when obtaining the summary) to include the variable with the lowest p-value and discard the one with the highest p-value.

My question is, is this procedure correct?

Thank you for your answers.

CJ.Learns.Math
  • 65
  • 6
  • Please explain what you mean by "correct." This is almost impossible to answer because both procedures are poor. – whuber Mar 03 '22 at 15:59
  • Either way is problematic for the same reasons mentioned when your [other question](https://stats.stackexchange.com/questions/566518/linear-regression-model-selection) got closed. Wald vs LR vs score in a setting other than stepwise variable selection (e.g., testing a pre-specified hypothesis) is a reasonable question, however. – Dave Mar 03 '22 at 16:02

0 Answers0