The Gradient Test: Another Likelihood-Based Test presents the latest on the gradient test, a large-sample test that was introduced in statistics literature by George R. Terrell in 2002. The test has been studied by several authors, is simply computed, and can be an interesting alternative to the classical large-sample tests, namely, the likelihood ratio (LR), Wald (W), and Rao score (S) tests.
Due to the large literature about the LR, W and S tests, the gradient test is not frequently used to test hypothesis. The book covers topics on the local power of the gradient test, the Bartlett-corrected gradient statistic, the gradient statistic under model misspecification, and the robust gradient-type bounded-influence test.
- Covers the background of the gradient statistic and the different models
- Discusses The Bartlett-corrected gradient statistic
- Explains the algorithm to compute the gradient-type statistic
Chapter 1 The gradient statistic
Chapter 2 The local power of the gradient test
Chapter 3 The Bartlett-corrected gradient statistic
Chapter 4 The gradient statistic under model misspecification
Chapter 5 The robust gradient-type bounded-influence test
Artur J. Lemonte is a professor at Federal University of Pernambuco, Department of Statistics, Recife/PE, Brazil. He works on higher order asymptotics, mathematical statistics, regression models, parametric inference, and distribution theory. In the last years, he has published more than 60 papers in refereed
statistical journals (most of them about the gradient test).