{"id":"arxiv/2601.00188","url":"https://arxiv.org/abs/2601.00188","tags":["stat.ME","math.ST"],"score":0,"title":"Decoupled Kemeny quasi-likelihoods for rank-based generalised linear models","domain":"stat.ME","source":"arxiv","authors":["Landon Hurley"],"reviews":0,"subject":"stat.ME","summary":"This paper introduces a unified profile quasi-likelihood framework for generalised linear models built on a topologically complete extension to Kendall's $τ$ with finite sample power and Type-I error rates. By establishing equivalence between continuous $L^2$ profile estimators and discrete Generalized Gibbs exponential families, we prove local asymptotic minimax efficiency over the copula tangent space. The framework yields monotone-affine invariant inference, sub-Gaussian concentration bounds, and exact finite-sample $t$-pivot tail majorization under arbitrary tie configurations. We…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-01-01","reactions":{}}