The Amelia package

Amelia is probably the newest multiple imputation package to be given to the R community. Amelia makes the assumption of multivariate normality, and in a moment we will see where this matters in terms of its use in R.

Tip

How important is the assumption of multivariate normality in imputation?

Many of the datasets we deal with in real life are not multivariate normally distributed, which might at first seem like a serious limitation to such an assumption. However, as the creators of Amelia point out, violations to this assumption usually don't make any practical difference. If this does bother an analyst, possible solutions include using the mice package that does not make this assumption or transforming the variables under study to be normally distributed.

We multiply impute data with the amelia command, which takes a number of arguments including the number of imputations to compute and descriptors for which variables are not continuous. By default, Amelia computes...