How does bagging affect linear model assumptions?

Question

Linear regression has assumptions. How does bagging affect model assumptions for linear regression?

Also, should you build a bagged linear model with correlated and statistically significant variables only? Does it matter if uncorrelated or statistically insignificant variables are used to build a bagged linear model?

How is bagging supposed to work for linear regression? Something like averaging multiple models as here: Can I merge multiple linear regressions into one regression?. But that won't reduce the variance. Or is this for some sort of regularisation and non-linear estimators? — Sextus Empiricus, Jul 09 '23 at 09:20

score 2 · Answer 1 · edited Jul 09 '23 at 09:19

linear regression has model assumptions. how does bagging effect model assumptions for linear regression?

The bagging procedures do not affect the model assumptions of linear regression. Bagging is a technique to reduce the variance of the model you use to bag (which in the case of linear regression is not super useful since it is a high bias low variance method to begin with).

also should you build a bagged linear model with correlated and statistically significant variables only?

No. Variable selection via p values is fraught with problems, as I and others have written about previously elsewhere on this site.

How does bagging affect linear model assumptions?

1 Answers1