scientific article; zbMATH DE number 3915460

From MaRDI portal
Publication:3690908





Let \(x_ 1,x_ 2,...\), be i.i.d. normal with mean \(\mu\) and variance \(\sigma^ 2\), both unknown. A class of sequential tests is studied for choosing one of two hypotheses concerning \(\mu\). These tests use the SPRT statistic \(Z_ n=x_ 1+...+x_ n\), and have boundaries \((-b_ ns^ 2_ n,b_ ns^ 2_ n)\) which converge to the SPRT boundaries \((- b\sigma^ 2,b\sigma^ 2)\) as \(n\to \infty.\) It is shown that the difference between the expected sample sizes of the proposed test and the SPRT is bounded as \(b\to \infty\) if \(b_ n=b(1+O(n^{-1}))\), and the asymptotic order of the error probability is found. Approximate expressions are derived for the error probability and the ASN, and these are compared with estimates obtained from simulations.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3690908)