1
$\begingroup$

Okays I have a quick question that I'm a little confused about.

During my glm model selection, the AIC from my most complicated model (model with say 7 predictors) and then the dredge function indicated which predictors were the most significant. I then ran the new model (with 4 predictors) to now only include those significant predictors. Now that i have ran this new model, 1 of those 4 previous predictors are now no longer significant. My question is, when plotting these effect plots, would I only plot the now 3 significant predictors, or would i still plot that 4th predictor as it was significant in the original beginning model?

Hopefully this isn't a complicated worded question, I am just confused about the change in significance

I understand dredging is not the best practice, however, my statistical abilities are not that great and I got it checked by my supervisor who said I would not be penalised for doing so.

$\endgroup$

1 Answer 1

1
$\begingroup$

To be consistent with your final model, only plot the effects for the variables that are actually in it.

However I must caution you that this kind of stepwise variable selection procedure is very, very bad and it is extremely depressing that your supervisor seems to think it is OK. Is there no underlying theory to which you can appeal ? You seems to be interested in inference rather than prediction, so you you to pay very careful attention to which variables are potential confounders, mediators, instruments and competing exposures.

$\endgroup$
1
  • $\begingroup$ Does this answer your question ? If so, please consider marking it as the accepted answer, and if not please let us know why. $\endgroup$ Aug 21, 2020 at 12:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.