Local asymptotic normality of ranks and covariates in transformation models (Q4355973)

\textit{L. Le Cam} and \textit{G. I. Yang} [Ann. Stat. 16, No. 2, 483-520 (1988; Zbl 0661.62016)] addressed broadly the following question: Given observations \(X^{(n)}= (X_{1n},\dots, X_{nn})\) distributed according to \(P^{(n)}_\theta\); \(\theta\in R^k\), such that the family of probability measures \(\{P^{(n)}_\theta\}\) has a locally asymptotically normal (LAN) structure at \(\theta_0\) and a statistic \(Y^{(n)}= g_n(X^{(n)})\): (i) When do the distributions of \(Y^{(n)}\) also have an LAN structure at \(\theta_0\)? (ii) When is there no loss in information about \(\theta\) in going from \(X^{(n)}\) to \(Y^{(n)}\)? We exhibit an important but rather complicated example to which the Le Cam-Yang methods may, after some work, be applied. For the entire collection see [Zbl 0861.00032].