Statistical Software for Data With Missing Values

In every large scale study, missing data is ubiquitous and inevitable. The overall cost incurredby the lost information is enormous, both financially, as cases are omitted due to incomplete data, andscientifically, as omitting data can lead to biased results. Research statisticians have developed highquality methods for estimating missing data, which can often enable the data analyst to use cases thatwould otherwise have to be dropped from the study. However, this research is not yet readily availableto the practicing analyst, because it is not embedded in software that is both accessible and easy to use.We will extend the S-PLUS statistical computing language to allow data analysts to input and manipulatemissing data models. We aim to support popular imputation methods, the EM algorithm, Markov ChainMonte Carlo, and multiple imputation methods. Moreover, we will provide an implementation that willpermit researchers to experiment more easily with new missing data methods, as the software will befully extensible.