Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: stata 11 mi and user written mim and ice commands


From   [email protected] (Yulia Marchenko, StataCorp LP)
To   Statalist <[email protected]>
Subject   Re: st: stata 11 mi and user written mim and ice commands
Date   Sun, 09 Aug 2009 14:13:23 -0500

David Airey <[email protected]> asked about the relationship between
the official command -mi- and user-written commands -ice- and -mim-:

> What is the relationship between Stata 11's -mi- commands and the user
> written -mim- and -ice- commands, especially -mi- vs -mim-? 

Multiple imputation consists of three phases: 1) imputation, 2) completed-data
analysis of multiply imputed data, and 3) pooling of individual analysis from
phase (2) using MI combination rules.

Stata 11's -mi- command provides subcommands to perform all three phases.
-mi- also provides a complete data management system for multiply imputed
data.

The user-written command -ice- performs phase (1) and the user-written command
-mim- performs phases (2) and (3).  -mim- also provides some capabilities for
manipulating multiply-imputed data.

There are several methods for performing the imputation phase. -ice- uses the
chained-equation method to impute multivariate missing data with an arbitrary
pattern of missing, whereas -mi impute- uses the NORM method.  See, for
example, http://www.stata.com/statalist/archive/2009-07/msg01107.html for more
discussion about the two methods.  -mi impute- also provides a number of
univariate imputation methods (-mi impute regress-, -mi impute logit-, etc.)
which are similar to the user-written command -uvis-.  When the pattern of
missingness is monotone, -mi impute monotone- can be used to impute
multivariate missing data similarly to -ice ..., monotone-.

-mi estimate- provides all of the estimation capabilities of -mim- except a
recently added Monte-Carlo error computation (Royston, Carlin, and White
2009).  It also allows estimation of nonlinear combinations of coefficients,
among other things.

Both -mim- and -mi estimate- provide testing of linear hypotheses of
coefficients.  -mi estimate- also implements small-sample adjustments for the
testing procedures and allows to test nonlinear hypotheses of coefficients.
-mim- provides a way to obtain predictions which is not available after -mi
estimate-.

All of the data management capabilities of -mim- and more are available in
-mi-.  Also, -mi impute- and -mi estimate- support factor variables.

We are currently working with Patrick Royston on integrating -ice- into -mi-.
Right now, you can use -mi import ice- and -mi export ice- to use the data
management and estimation capabilities of -mi- with imputed data obtained from
-ice-.


Reference:

P. Royston, J. B. Carlin, and I. R. White. 2009. Multiple imputation of
missing values: New features for -mim-. Stata Journal 9(2): 252-264.  


Best,

--Yulia
[email protected]

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index