Metternich, N; Hollenbach, FM; Bojinov, I; Minhas, S; Ward, MD; Volfovsky, A; (2019) Multiple Imputation Using Gaussian Copulas. Sociological Methods and Research 10.1177/0049124118799381. (In press).

Preview

Text 1411.0647.pdf - Accepted Version Download (649kB) | Preview

Abstract

Missing observations are pervasive throughout observational research, especially in the social sciences. Despite multiple approaches to dealing adequately with missing data, many scholars still rely on list-wise deletion. In this article, we present a simple to use approach to multiple imputation. We show that using Gaussian copulas for multiple imputation allows scholars to attain estimation results that have good coverage and small bias. Using simulated as well as observational data from published social science research we compare imputation via Gaussian copulas with two other widely used imputation methods: MICE and Amelia II. The three approaches perform relatively similarly. Importantly, however, imputation via the Gaussian copula is simple and does not require the researcher to undertake any transformation of the data or specification of distributional assumptions for individual variables but returns a valid posterior density of the imputed data.

Download activity - last month

Export as CSVJSONXML

Download activity - last 12 months

Export as CSVJSONXML

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item