The National Survey on Drug Use and Health (NSDUH) is the primary source of information on drug substance use in the U.S. Since 1999, the Predictive Mean Neighborhoods (PMN) procedure has been used to impute missing values for many of the analytical variables. This method is a combination of two commonly used imputation methods: a nearest-neighbor hot deck and a modification of Rubin's predictive mean matching method. Although PMN has many practical advantages, it has not been formally evaluated. We propose a simple simulation to evaluate PMN. Using only complete data cases, we will induce random patterns of missingness in the data for selected outcome variables. Imputations will then be conducted using PMN and weighted and unweighted sequential hot decks. This process of inducing missingness and imputing missing values will be repeated multiple times. The imputed values using PMN and the hot deck methods will then be compared with the true values that were found in the complete data, across the repeated iterations. In particular, we will compare the number of matches between the two methods, as well as comparing statistics derived from the data, such as drug prevalence estimates.
