[R] Problems to obtain standardized betas in multiply-imputed data

[Paul Johnson]

Greetings.
I would adjust approach  to calculate standardized estimates for each
imputed set. Then summarize them . The way you are doing it here implies
that standardization concept applies to model list, which seems doubtful.
The empirical std. dev. of the variables differs among imputed data sets,
after all.

I suppose I mean to say lm.beta is not intended to receive a list of
regressions. Put standardization in the with() work done on each imputed
set. I suspect it is as easy as putting lm.beta in there. If there is
trouble, I have a standardize function in the rockchalk package. Unlike
lm.beta, it actually standardizes variables and runs regression. lm.beta
resales coefficients instead.

Paul Johnson
University of Kansas

On Wed, Sep 26, 2018, 5:03 AM CHATTON Anne via R-help <r-help using r-project.org>
wrote:

> Dear all,
>
> I am having problems in obtaining standardized betas on a multiply-imputed
> data set. Here are the codes I used :
> imp = mice(data, 5, maxit=10, seed=42, print=FALSE)
> FitImp <- with(imp,lm(y ~ x1 + x2 + x3))
> Up to here everything is fine. But when I ask for the standardized
> coefficients of the multiply-imputed regressions using this command :
> sdBeta <- lm.beta(FitImp)
> I get the following error message:
> Error in b * sx : argument non numérique pour un opérateur binaire
>
> Can anyone help me with this please?
>
> Anne
>
> ______________________________________________
> R-help using r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

	[[alternative HTML version deleted]]