Asymptotic sufficiency. II: Truncated cases (Q799043)

[For part I see the preceding review, Zbl 0547.62002.] Higher order asymptotic sufficiency is investigated in the iid translation parameter case generated by some density p(x) which vanishes on (-\(\infty,0)\) and is positive on (0,\(\infty)\). It is assumed that \(p(x)\sim ax^{1+b}\) as \(x\to 0+\), where a is positive and b is nonnegative. For this situation, asymptotic sufficiency of maximum likelihood estimators has been investigated by \textit{H. Mita}, Tokyo J. Math. 2, 323-335 (1979; Zbl 0443.62021). Higher order asymptotic sufficiency is derived for statistics containing the maximum likelihood estimator and derivatives of the log-likelihood function up to order k. It is shown that these statistics are asymptotically sufficient of order o((log n)\({}^{-r})\) for every \(r<(k+1)/(k+3)\) if \(b=0\) and the Fisher information is infinite; if \(0<b\leq k(k+3)\), then the corresponding order is \(O(n^{-r})\) for every \(r<b(k+1)/(2(k+b+3))\); if \(b>k(k+3)\), then the order is \(o(n^{-k/2})\).