The usual Reinterview program consists reinterviewing a set of respondents and the data of the two interviews yields such measures as: Index of Inconsistency, Simple Response Variance, Gross Difference Rate, and Net Difference Rate etc. However the underlying model has a potential weakness namely that the respondent may be influenced by their first response. This can be effectively eliminated by splitting the sample into two segments on whether they remembered initial response or not. The resulting two tables corresponding to the split samples yield similar measures as above but they bear some interest in relationships. In this paper we derive these relations.
Keywords: Simple Response Variance; Correlated Response Variance; Response Bias
Biography: Dhiren Ghosh is a Sampling Statistician at Synectics for Management Decisions Incarporated. He has more than 50 years of experience in Sample Survey and has been a Statistian to hundreds of projects. He was originally trained in the Indian Statistical Institute under P.C. Mahalanobis and he is fellow of the Royal Statistical Society.