This work compares two mean estimators, MV and MKL, which incorporate information about a known quantile. MV minimizes variance and MKL minimizes Kulback-Leibler divergence. Both estimators are asymptotically equivalent and normally distributed but dier at nite sample sizes. Monte-Carlo simulation studies show that MV has higher mean squared error than MKL in the majority of simulated scenarios. Authors recommend using MKL when a quantile of an underlying distribution is known.
Пятая Международная конференция по стохастическим методам (МКСМ-5) : материалы Международной научной конференции, Россия, Москва, 23-27 ноября 2020 г.. М., 2020. С. 236-240