Assessing the Effectiveness of a Stochastic Regression Imputation Method for Ordered Categorical Data
Isabella Sulis, Mariano Porcu · RePEc: Research Papers in Economics · 2008
The main aim of this paper is to describe a workable method based on stochastic regression and multiple imputation analysis (MISR) to recover for missingness in surveys where multi-item Likert-type scale are used to measure a latent attribute (namely, the quality of university teaching). A simulation analysis has been carried out and results have been compared in terms of bias and efficiency with other missing data handling methods, specifically - Complete Cases Analysis (CCA) and Multiple Imputation by Chained Equations (MICE). The authors provide also functions (implemented in R language) to apply the procedure to a matrix of ordered categorical items. Functions described allow - (i) to simulate missing data at random and completely at random; (ii) to replicate the simulation study presented in this work in order to assess the accuracy in distribution and in estimation of a multiple imputation procedure.