A lower bound for the bayes risk in the sequential case (Q4240731)

Let \(X_1,X_2, \dots, X_n,\dots\) be independent and identically distributed random variables with a density function \(f(x,\theta)\) relative to a \(\sigma\)-finite measure \(\mu\), where \(\theta\in \Omega\subset {\mathcal R}^1\). The author considers the problem of sequential estimation of \(g(\theta)\) under the loss function \(L=(\widehat g_\tau- g(\theta))^2 +(c\tau/ I(\theta))\), where \(\tau\) is a stopping time, \(\widehat g_\tau\) is an estimator, \(c\) is unit cost of sampling per observation, and \(I(\theta)\) is the Fisher information. Under suitable regularity conditions on \(f(x,\theta)\), \(\tau\), and the density \(\pi (\theta)\), the author obtains a lower bound for the Bayes risk for the sequential estimation procedure. The given result is an extension of the work of \textit{M. Alvo} [Ann. Stat. 5, 955--968 (1977; Zbl 0368.62061)].