APPROXIMATE AND GENERALIZED APPROACHES TO CONFIDENCE INTERVALS ON A VARIANCE COMPONENT

We consider a construction of confidence intervals on a variance component in two-component mixed linear models. Firstly, some existing approximate intervals and their properties are discussed, then the problem is considered from the viewpoint of generalized inference and the possibility of using generalized intervals as confidence intervals (in our setting) is examined. In particular, we focus on intervals obtained by a generalized pivot, firstly proposed by [D. J. Park, R. K. Burdick: Performance of confidence intervals in regression models with unbalanced one-fold nested error structures, Comm. Statist. Simulation Comput. 32 (2003), 717-732].