Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: multiple imputation + transformation of parameters

From   Constantine Daskalakis <>
Subject   Re: st: multiple imputation + transformation of parameters
Date   Thu, 13 May 2004 11:24:14 -0400

At 05:55 AM 5/13/2004, Marcello Pagano wrote:
Michael is having problems posting this:


-------- Original Message --------

  I want to report confidence intervals for odds ratios.  It's clear
that -MIFIT- is working with the underlying betas even when one uses the
logistic command.  What STATA reports as the MI estimate for the odds
ratio is the exponentiated value of the betas averaged across the

   It makes more sense to me to transform the betas before the
averaging.  My sense is that one should do all steps in the analyses
before implementing Rubin's rules.

   Does this sound right?  Can anyone make a counter-argument?

I do not think so.
The estimands are the betas in the regression. Therefore, imputation should be on the betas.
Also, the distribution of the beta-hats is probably going faster to normality than that of exp(beta-hat). It is the same reason why we do asymptotic CIs on the log odds ratio scale (beta-hat plus/minus z*se[beta-hat]) and *then* exponentiate the limits, rather than doing the calculation directly on the odds ratio scale.

The documents accompanying this transmission may contain confidential health or business information. This information is intended for the use of the individual or entity named above. If you have received this information in error, please notify the sender immediately and arrange for the return or destruction of these documents.

Constantine Daskalakis, ScD
Assistant Professor,
Biostatistics Section, Thomas Jefferson University,
211 S. 9th St. #602, Philadelphia, PA 19107
Tel: 215-955-5695
Fax: 215-503-3804
* For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index