A Generalization of an SPRT for the Correlation Coefficient

Abstract

The sequential probability ratio tests (SPRT's) for the correlation coefficient, ρ, of a bivariate normal distribution proposed by Kowalski [3] and Pradhan and Sathe [4] reduce to SPRT's of a binomial parameter. These tests, respectively, require the sampling of one and two pairs of observations at each step of the sequential procedure. This article generalizes their SPRT's to that of sampling K (K = 1, 2, 3, …) pairs of observations at each step and compares the tests thereby generated.