Nick has a paper in the American Statistician warning about bias in multiple imputation arising from rounding data imputed under a normal assumption. One example where you might run afoul of this is if the data are truly dichotomous or count variables…