An Improved Integral-Typed Goodness-of-Fit Test Under Crossing Survival Functions

• Activity Number: 414 - Risk Modeling and Regression Techniques
• Type: Contributed
• Date/Time: Thursday, August 12, 2021 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Hsin-wen Chang*
• Companies: Academia Sinica
• Keywords: bootstrap; empirical distribution function; stochastic process

Abstract

Nonparametric goodness-of-fit testing has been applied widely in practice, yet it can suffer a loss of power when there is a crossing between the distribution functions being compared. To deal with this issue, in this paper develop an improved test statistic that takes into account crossing. The proposed test is based on the classical Anderson--Darling statistic. Our simulations show the proposed statistics outperform traditional statistics in the setting of crossing distribution functions.