Comparing Non-Nested Linear Models

摘要: Abstract I consider the usual linear-model situation, except that there are two possible linear subspaces may contain true mean vector, and neither of is nested within other. Approximate confidence intervals developed for difference in squared error (MSE) prediction using models, not assuming either model necessarily correct. The based on parametric bootstrap methods, applied to Mallows's Cp estimate MSE. This approach shown relate closely Hotelling's test comparing simple regressions. In simplest case problem equivalent finding a interval product means independent normal observations, each with variance one.