Studentising non-parametric correlation estimators

Abstract: Studentisation upon rank-based linear estimators is generally considered an unnecessary topic, due to the domain restriction upon $S_{n}$, which exhibits constant variance. This assertion is functionally inconsistent with general analytic practice though. We introduce a general unbiased and minimum variance estimator upon the Beta-Binomially distributed Kemeny Hilbert space, which allows for permutation ties to exist and be uniquely measured. As individual permutation samples now exhibit unique random variance, a sample dependent variance estimator must now be introduced into the linear model. We derive and prove the Slutsky conditions to enable $t_{\nu}$-distributed Wald test statistics to be constructed, while stably exhibiting Gauss-Markov properties upon finite samples. Simulations demonstrate convergent decisions upon the two orthonormal Slutsky corrected Wald test statistics, verifying the projective geometric duality which exists upon the affine-linear Kemeny metric.