A Monte Carlo analysis of missing data techniques in a HRM setting

Abstract

Researchers have examined various techniques to solve the problem of missing data. Simple techniques have included listwise deletion, pairwise deletion, mean substitution, regression imputation and hot-deck imputation. Past research suggests that regression imputation and pairwise deletion generally result in less dispersion around true score values while listwise deletion results in more dispersion around true scores. Unfortunately, this research spent much less time examining whether the various techniques lead to overestimation or underestimation of the true values of various statistics. The present study utilized a Monte Carlo Analysis to simulate an HRM research setting to evaluate missing data techniques. Pairwise deletion resulted in the least dispersion around true scores and least average error of any missing data technique for calculating correlations. Implications for use of these techniques and future missing data research were explored.