How to Deal with Missing Data and Galton’s Problem in Cross-Cultural Survey Research: A Primer for R

Abstract

Multiple imputation (MI) has become the preferred method for dealing with missing data in survey research. MI involves three steps: creating m multiply imputed complete datasets; estimating models on each of the m datasets using any standard statistical procedure; combining the resulting multiple estimates of each statistic of interest. This paper provides R programs for MI, and offers some advice for employing MI with data drawn from the Standard Cross-Cultural Sample (SCCS). A second set of R programs combines estimates from the m imputed data sets, and also deals with the problem of network autocorrelation effects, i.e., Galton’s Problem or the non-independence of cases, using two-stage instrumental variables (IV) regression. The objective of the paper is to provide programs, advice and explanations that will help researchers employing cross-cultural survey data, especially the SCCS, to deal with the twin problems of missing data and network autocorrelation effects, using the open source statistical package R. The paper is intended to complement a recent suite of publications by Dow and Eff where both theoretical and empirical issues underlying these two problems are discussed in detail.