Imputation for Missing Data: A Simulation Study
by
Girish Kumar Jha
Indian Agricultural Statistics Research Institute,Library Avenue, Pusa, New Delhi -12, India

Modern complex surveys typically collect responses to a large number of items for each sampled unit and are usually subject to non-response errors. Imputation is a very common technique for handling item non-response. An imputation method is defined as a procedure that imputes a value for each missing item. In this study, an attempt has been made to empirically investigate the performance of existing imputation methods under conditional framework where the groups have been formed on the basis of ancillary statistics. Recent progress in the field of artificial neural networks (ANNs) indicates that imputation can be reformulated as a neural network problem. The possibility of using ANNs based techniques for the purpose of imputation has also been discussed in the paper.

Date received: November 30, 2004