Estadística de gradiente: una opción para realizar pruebas de hipótesis en escenarios de tamaño de muestra pequeño
PublicadorInternational Journal of Applied Engineering Research
The gradient statistic, recently proposed in the literature, has gained attention on statistical practitioners due to the fact that it is a competitive alternative for the traditional statistics (likelihood ratio, Wald, Score) for performing hypothesis testing in parametric models with the special property of being easily computable. In this work, we present an exhaustive Monte Carlo simulation study with the four test statistics to assess its performance in a controlled scenario based on the exponential distribution. The obtained results suggest that the gradient test statistics is, indeed, a competitive option for the "Holy trinity" of statistical inference (Likelihood ratio, Score and Wald test statistics).