A Markov Chain Monte Carlo algorithm for multiple imputation of missing values.
Vallinayagam, V.; Senthamaraikannan, K.; Venkatesan, P.
Book published by Allied Publishers P Ltd. on Mathematics, Computing and Modeling. Ed. P. Balasubramaniam and R. Uthayakumar. 2007; 257-264.
Abstract: The imputation methods for missing values allow valid estimates for the variance of the estimates to be calculated using standard complete data methods. Multiple imputation procedure replaces each missing value by a vector of imputed values and creates complete data sets. These methods depend on the missing patterns in the data. In this paper we present the methods for multiple imputations under various missing patterns. In particular this work proposes Markov Chain Monte Carlo (MCMC) method based imputation for arbitrarily missing patterns and compares the results with regression and propensity score based methods. The work deals with new algorithms for arbitrarily missing data. The MCMC methodology to missing data is also compared with other algorithms and its advantages are established. The results are presented and discussed.
Back to List of publications / Home