• Poster

A Simple Evaluation of the Predictive Mean Neighborhoods Imputation Procedure

Citation

Grau, E. A., Frechtel, P. A., & Odom, D. M. (2004, August). A Simple Evaluation of the Predictive Mean Neighborhoods Imputation Procedure. Presented at , .

Abstract

The National Survey on Drug Use and Health (NSDUH) is the primarysource of information on drug substance use in the U.S. Since 1999, thePredictive Mean Neighborhoods (PMN) procedure has been used to imputemissing values for many of the analytical variables. This method is acombination of two commonly used imputation methods: a nearest-neighborhot deck and a modification of Rubin's predictive mean matching method.Although PMN has many practical advantages, it has not been formallyevaluated. We propose a simple simulation to evaluate PMN. Using onlycomplete data cases, we will induce random patterns of missingness inthe data for selected outcome variables. Imputations will then beconducted using PMN and weighted and unweighted sequential hot decks.This process of inducing missingness and imputing missing values willbe repeated multiple times. The imputed values using PMN and the hotdeck methods will then be compared with the true values that were foundin the complete data, across the repeated iterations. In particular, wewill compare the number of matches between the two methods, as well ascomparing statistics derived from the data, such as drug prevalenceestimates.