Conical tests: Observed levels of significance and confidence regions

From MaRDI portal





In this paper the authors consider the regression model \(Y=X\beta +\sigma e\), where X is an \(n\times p\) matrix of rank p and errors are i.i.d. rv's distributed according to some d.f. F. Conical tests of significance for hypotheses about the parameters and conical confidence regions for them, which involve a reasonable amount of calculations, are proposed and studied. The proposed conical tests provide confidence regions, giving the confidence bound at a chosen level of significance in any direction from a central 0-confidence point. The approach is illustrated on several examples, both when the distribution of errors is normal (\(\sigma\) known, resp. \(\sigma\) unknown) and when it is nonnormal.











This page was built for publication: Conical tests: Observed levels of significance and confidence regions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3713383)