A sequential multinomial selection procedure with elimination
The authors discuss a multinomial procedure for selecting the best of a number of alternatives. Suppose \(t\) categories have associated (unknown) probabilities \(p_1,p_2,\ldots,p_{t}\), \(\sum_{i=1}^{t}p_{i}=1.\) It is possible to observe a random sample of \(n\) multinomial observations, \({\mathbf x}_{m}= (x_{1m},x_{2m},\ldots,x_{tm})\), \(1\leq m \leq n.\) Here \(x_{im}=1[0]\) if event \(i\) occurs [does not occur] on the \(m\)-th observation, \(\sum\limits_{i=1}^{t}x_{im}=1\), \(1\leq m \leq n.\) The problem is to determine which of the competing categories is associated with the largest unknown multinomial cell probability. The authors consider the following decision rule: calculate the sample sums \(y_{in}= \sum\limits_{j=1}^{n}x_{ij}\), \((1\leq i \leq t)\) and select the category that yields the largest sample sums \(\max\limits_{1\leq i \leq t}y_{in}\), as the one associated with \(\max p_{i}.\) The authors explain how many observations must be taken in order to satisfy the requirement: the probability of correct selection is greater then \(P_{\ast}\) whenever \(p_{[t]}/p_{[t-1]} \geq\theta^{\ast}\), where \(p_{[i]}\), \(i=1,2,\ldots,t\) are the ordered \(p_{i}\)'s, and \(P_{\ast}\) and \(\theta_{\ast}\) are user-specified constants. Moreover, the authors explain how one can reduce the number of observations. For the entire collection see [Zbl 0990.62506].
- A sequential elimination procedure for choosing the best population(s) based on multiple testing
- Selecting the best of k multinomial parameter estimation procedures using sprt
- Multinomial selection in the presence of infinite alternatives
- Optimal selection of the most probable multinomial alternative
- Closed sequential procedures for selecting the multinomial events which have the largest probabilities
- Selecting the highest of three binomial probabilities
- Sequential Procedures for Selecting the Most Probable Multinomial Cell When a Nuisance Cell Is Present
- A restricted multinomial hybrid selection procedure
This page was built for publication: A sequential multinomial selection procedure with elimination
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3843177)
