scientific article; zbMATH DE number 4123079

From MaRDI portal
Publication:4203646





[For the entire collection see Zbl 0637.00008.] Let \(\Phi\) be a set of probability densities on a measurable space (\({\mathcal X},{\mathcal A})\) and \(X_ 1,...,X_ n\) be a random sample with unknown density \(f\in \Phi\). Let the null hypothesis \(H_ 0\) be \(f\in \Phi_ 0\) with \(\Phi_ 0\subset \Phi\). Given a measure \(d(f,\Phi_ 0)\) of distance for a density \(f\in \Phi\) from the set \(\Phi_ 0\) and given a positive number \(\rho_ N\) the author considers as alternative hypothesis \(H_ 1\) the set of densities \[ \Phi_ N=\{f\in \Phi:\quad d(f,\Phi_ 0)\geq \rho_ N\}. \] The test problem is considered from the minimax point of view. For a given test \(\psi_ N\) let \(\gamma_ N(\psi_ N)\) be the sum of the first and second kind error probabilities under \(\psi_ N\). Let \(\gamma_ N\) be the minimum value of \(\gamma_ N(\psi_ N)\). The number \(\gamma_ N\) characterizes the minimax discernability of \(H_ 0\) and \(H_ 1.\) The case of a nonparametric \(\Phi\) is considered and an asymptotic version (N\(\to \infty)\) of the minimax hypotheses testing problem is investigated.




Cited in
(41)








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 Q4203646)