scientific article; zbMATH DE number 19002
[For the entire collection see Zbl 0729.00014.] The author considers the following situation: Let \(X_ 1,\dots,X_{n_ 1}\sim B(1,P_ 1)\) and \(Y_ 1,\dots,Y_{n_ 2}\sim B(1,P_ 2)\) be independent Bernoulli trials, where \(P\in\Omega= (0,1)*(0,1)\). Let \(\theta=P_ 2- P_ 1\). To estimate \(\theta\) the maximum likelihood estimator is used. The problem consists of determining how to allocate patients to treatment groups so that, on average, the loss measured by \[ L_ n(P,\widehat \theta_{n_ 1,n_ 2})= n\Bigl[\bigl\{\sqrt n(\theta-\widehat\theta_{p_ n})\bigr\}^ 2+ p_ n(1- P_ 1)+ (1- p_ n)(1-P_ 2)\Bigr], \] where \(p_ n=n_ 1/n\) and \(\widehat\theta_{p_ n}=\widehat\theta_{n_ 1,n_ 2}\), will be as small as possible. Several rather classical methods using dynamic programming and backward induction are used to find optimal solutions.
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 Q3976427)