Imputation techniques in regression analysis: looking closely at their implementation

摘要: A problem which frequently arises in regression analysis is the presence of missing values on explanatory variables. Imputation a time-honoured approach to tackling it, since graphical exploration properties statistical model requires complete data matrix. This article examines performance five imputation techniques two used implementation procedures. Specifically, imputed based both response and variables (type I) are contrasted with those only II). Monte Carlo results indicate that type I procedure may give spurious impression high precision especially as proportion increases. But II, overestimation residual mean square error arise. Several matrices correlation coefficients an illustrative real example given.