Spotfire S+ Missing Data library

The Spotfire S+ "missing" library contains software for handling missing
data using a model-based approach. The EM and data augmentation
(Gibbs sampler) algorithms are used to fit the models.

The data augmentation algorithms also produce multiple imputations.

Multiple imputation creates several plausible values for each missing
value and allows the analyst to estimate additional uncertainty due
to missing data.  Multiple completed data sets arise from substituting
each set of imputations for the missing values.  The library includes
capabilities to easily perform complete data analyses on the multiple
completed data sets, and combine inferences to reflect the additional
uncertainty due to missing data.

Users may choose to write their own routines for creating multiple
imputations, yet still want to use the software to perform multiple
complete data analyses and consolidate results.  This is possible as
long as they first create the appropriate Spotfire S+ objects.

The software is based on algorithms described by Little and Rubin
(1987) and Schafer (1997).  Joe Schafer devised the data augmentation
algorithms.  The code builds upon code written by Joe Schafer.

The Insightful Corp. Research developers include Jim Schimert, Douglas
B. Clarkson, Chris Fraley, Tim Hesterberg, and Shan Jin.  Special
thanks to Stephen Kaluzny for helping with builds.  This work was
supported by NIH SBIR grants R43RR0925401, R44CA65147-02, 03.


To use the library, first load it, e.g. from the command line use:

> library(missing)

To summarize and view patterns of missing data, see functions:
        miss
        numberMissing

To estimate parameters using EM or data augmentation, see functions:
        mdGauss
        mdLoglin
        mdCgm

To create multiple imputations using data augmentation, see functions:
        impGauss
        impLoglin
        impCgm

To convert your own multiple imputations into the appropriate Spotfire S+
objects for use with this library, see functions:
        miVariable
        miList

To perform analyses in parallel on multiple imputations, see functions:
        miApply
        miEval

To consolidate numerical results and standard errors from statistical
analyses from multiple imputations, see functions:
        miMean         miChiSquareTest
        miMeanSE       miLikelihoodTest
        miSummary      miFTest

The help files for these functions contain examples and references
to additional functions.


Extensive documentation is provided in the manual, file
        Missing.pdf
located in directory
        `Splus SHOME`/library/missing
(replace Splus with the command you use to start this version of Spotfire S+).
