scientific article; zbMATH DE number 3990712
[For the entire collection see Zbl 0598.00013.] This article centers around the construction of upper confidence bounds (UBC) for a response probability \(\pi (x_ 0)\) from values of a dose- response curve \(\pi\) (x), \(x\geq x_ 0\), in the analysis of quantal bioassay experiments. This is achieved for a unihypermodal \(\pi\) (x) when \(\pi\) (0) and its hypermode H is known, as well as when an UBC of \(\pi\) (x) is given for \(x_ 0\leq x\leq H\). Thereby, \(\pi\) is called unihypermodal with hypermode H if the response-to-dose ratio \(\rho (x)=\pi (x)/x\) is unimodal in x with mode H. Within the class of sigmoidal curves \(H\geq M\), if M denotes the inflection point of \(\pi\). Two classes of hypothesis tests about the location of the hypermode H are derived: one based on a likelihood ratio statistic and another based on a weighted sum of constrained contrasts (linear test statistic). The hypothesis tests lead to a lower confidence bound for H, which in turn is used to construct an UBC for \(\pi (x_ 0)\). Both, deficiencies of the tests as well as practical utilities are discussed. The computation of maximum likelihood estimates of response probabilities of a hypermodal \(\pi\) is obtained by a maximum violator algorithm with guaranteed convergence.
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 Q3753359)