Corrected score tests for exponential censored data

Abstract

In this paper we discuss the issue of improving score statistics for exponential censored data. We give a general formula to compute corrected score statistics in exponential regression models. The formula derived is simple enough to be used analytically in order to obtain closed-form Bartlett-type corrections to improve score statistics in several special cases. The formula has also advantages for numerical purposes because it requires only simple operations on matrices. We show by Monte Carlo simulations that the corrected score test seems to improve over the usual score test.