On difference of means of correlated variates with incomplete data on both responses

Abstract

A class of estimators is proposed for the difference of means when sampling from a biv ariate normal distribution with means w, and u2, variances o\ and <xf, and correlation coefficient p, where some observations on either of the variables are missing. The new estimator from this class is used to obtain some test statistics for testing the equality of means, Some additional test statisticsare also proposed and the empirical powers of all statistics arecomputed for different values of <xf, o\ and p. These computationssupport the use of some of the new teststatistics

Keywords:

• Combination of independent tests
• empirical size and power
• equality of means
• Fisher'sme thod
• incomplete data
• linear combination of independent variates
• paired t test
• unbiased estimator