Need help: how to do bonferroni adjustment in the multiple linear regression

[inky]

Hi! I'm inky. This is the first participation in the discussion.

Actually, I need your help to resolve my struggle; I have struggled with the reviewer's request regarding my manuscript, especially statistical method.

I used the multiple linear regression analysis (PROC REG) to evaluate the association between a biomarker (Y: long-transformed continuous variable)

and a factor score (X: continuous variable) after adjustment of age, sex, socioeconomic status, and other covariables.

The association was weakly significant (p-value <0.05)

The reviewer requested me to present p-value after Bonferroni adjustment in this model.

In my understanding, Bonferroni adjustment is relevant in multiple comparison for a categorical variable.

But, the reviewer said that "The Bonferroni adjustment is applicable whenever multiple p-values are considered regardless of which
statistical models were used to derive the p-values."

Is this correct for continuous variables in the multiple linear regression model?

If so, how to do Bonferroni adjustment in PROC REG?

Please get me out of this difficulty.

[PGStats · Accepted solution]

The need to correct for test multiplicity is mostly a matter of opinion (discuss it with your editor) but if you need to do so for any set of independent p-values, look at proc multtest with option inpvalues=

PG

[inky]

Dear PG

Thank you a lot. I did run PROC MULTTEST as you suggested. p-value was not significant.

[PaigeMiller]

Please understand that the Bonferroni adjustment gives you very conservative results compared to other multiple comparison methods. In other words, you are much less likely to find statistically significant results.

Furthermore, I agree with , there's no global agreement that you should use Bonferroni in this situation, and most statistical literature doesn't use them in this situation; most (if not all) of the times that I have seen Bonferroni used, it is for comparing means and not for slopes in a multiple regression.

[inky]

Dear Mr. Miller

Thank you a lot for your opinion, which is very helpful for me to understand Bonferroni and to resolve my struggle.

Inky