SEQUENTIAL EXPERIMENTS IN BIOASSAY.
Abstract
Five sequential and two non-sequential designs for estimating the median lethal dose m in the biological assay problem when quantal responses are studied and compared. Three versions of the stochasticapproximation process of Robbins and Monro (1951) and two versions of the up-and-down method of Brownlee, Hodges and Rosenblatt (1953) are compared with non-sequential designs using minimum transform chisquare estimates of Berkson (1956) and Spearman-Karber estimates described by Cornfield and Mantel (1950). Exact mean square errors of the estimate of m are computed for several initial doses, step sizes, arrangements of animals in trials and forms of assumed dose-response curves. The sequential designs appear to have smaller mean square error than the non-sequential ones under almost all conditions. If the experimenter is sure that he knows m within 3 sigma, all the sequential designs perform approximately equally well. If he is less certain of m, the delayed Robbins-Monro process of Cochran and Davis (1965) or the delayed up-and-down method are clearly more accurate than the others. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 30, 1965
- Accession Number
- AD0634864
Entities
People
- Miles Davis
Organizations
- Harvard University