On partial nonresponse situations:the hot deck imputationmethod

Dealing with data files statisticians often have to consider the problem of missing data due both to unit nonresponse (complete nonresponse) and item nonresponse (partial nonresponse). As it regards item nonresponse (see e.g. Kalton and Kasprzyk (1986)) different forms of imputation (deterministic or stochastic) are often used (see e.g. Little and Rubin (1987), Cicchitelli, Herzel, Montanari (1992)).A very well known technique used to fill in values for these missing values is the hot deck imputation (see Ford (1983)). In this paper we consider the problem of variance estimation from imputed survey data under the hot deck method.