B0758
Title: The modified ($\Phi$, $\alpha$)-power divergence family: The zero count case
Authors: Christos Meselidis - University of the Aegean (Greece) [presenting]
Alexandros Karagrigoriou - University of The Aegean (Greece)
Abstract: Divergence measures play a crucial role in statistical inference. They can be used in the construction of tests of fit as well as for estimation purposes. In the presence of zero frequency cells, a modified version of the ($\Phi$, $\alpha$) power divergence family, which produces robust estimators that are more efficient than the classical ones, can be utilized. These estimators are then used in the proposed test statistic that involves four indices, through which the stability of the test is achieved in the case of contaminated data. All the aforementioned notions are presented through an extensive simulation study. Regarding the estimation procedure, conclusions are based on the mean squared error, while for the test statistic, both the size and the power of the test are taken into consideration.
