Comparison of confidence intervals for correlation coefficients based on incomplete monotone samples and those based on listwise deletion

Abstract

Inferential procedures for estimating and comparing normal correlation coefficients based on incomplete samples with a monotone missing pattern are considered. The procedures are based on the generalized variable (GV) approach. It is shown that the GV methods based on complete or incomplete samples are exact for estimating or testing a simple correlation coefficient. Procedures based on incomplete samples for comparing two overlapping dependent correlation coefficients are also proposed. For both problems, Monte Carlo simulation studies indicate that the inference based on incomplete samples and those based on samples after listwise or pairwise deletion are similar, and the loss of efficiency by ignoring additional data is not appreciable. The proposed GV approach is simple, and it can be readily extended to other problems such as the one of estimating two non-overlapping dependent correlations. The results are illustrated using two examples.