| Summary: | Part of a special section on score test oddities. The score test, as a tool, is commonly used in statistical inference due to its simplicity, elegance, and convenience, but it does present a number of issues. These issues can be explored along four dimensions: unconstrained versus constrained parameter spaces; use of observed versus expected information; rejection probabilities carried out under the null hypothesis, a correctly specified alternative, or a misspecified alternative; and asymptotic versus small-sample behavior. The writers consider the first two dimensions in detail.
|
|---|