WinMICE by Gert Jacobusse is a stand-alone program under Windows that implements imputation on the linear mixed model. Needs Windows.

ICE by Patrick Royston. Elegant STATA implementation of MICE. See his presentation for more details. Needs STATA.

IVEWARE by Raghunathan, Solenberger and John Van Hoewyk is a SAS-based application for creating multiple imputations. Similar to MICE and quite flexible. Imputations can be constrained. Both for Windows and UNIX. Needs SAS.

Missing Data Library in S-Plus 6. Based upon the work of Joseph L. Schafer, it features Gaussian, Loglinear and Conditional Gaussian. Performing multiple complete data analysis after multiple imputation, and consolidating results, is simplified by using the library. More info. Standard in S-Plus.

The S-Plus Hmisc library by Frank Harrell contains a number of functions for imputing missing data (both single and multiple), as well as facilities for analysis and pooling. Needs S-Plus 3.0 or higher. Now in standard distribution of S-Plus.

SOLAS for Missing Data Analysis 3.0 is a commercial Windows program by Statistical Solutions Limited. Version 3.0 offers new methods for multiple imputation, primarily based on Rubin's Chapter 5, as well as a scripting facility Stand-alone, pretty interface.

SAS announced two new procedures: PROC MI and PROC MIANALYZE. Current version is V8.2 and documentation is now complete. There is also a SUGI paper that gives a preview (.pdf, 90Kb).

NORM, CAT, MIX and PAN is software for multivariate imputed by Joseph L. Schafer. NORM uses a normal model. CAT uses a loglinear model for categorical data. MIX relies on the general location model for mixed categorical and continuous data. PAN is geared toward panel data. S-PLUS 3.3 and 4.0 and stand-alone Windows software is available.

AMELIA: A Program for Missing Data by James Honaker, Anne Joseph, Gary King, Kenneth Scheve and Naunihal Singh uses the EMis method, a speedy version of NORM. It comes in versions for DOS and Gauss, and it's unique for time series.

EMCOV by John W. Graham and Scott M. Hofer runs on DOS and UNIX and uses the EM algorithm in the presence of missing data. It can also impute missing values. Contains additional NORM utilities. Stand-alone.

Steve Gregorich has written SAS macros for estimating the ML covariances matrix and mean vector by the EM algorithm.

Paul Allison's SAS macros are forerunners of PROC MI and PROC MIANALYZE.

For special applications: Check out MISTRESS by Stef van Buuren and Jan van Rijckevorsel. This method creates single imputations that maximise the internal consistency in categorical data.

impute.lsp by Jason Bond implements multivariate normal imputation in X-LISP. Needs X-LISP.

Multiple imputation code in SAS as done in Little and Yau (1996) for intent-to-treat analysis of repeated measures data with drop-outs.