CMStatistics 2021: Start Registration
View Submission - CMStatistics
B0665
Title: Specification tests for selection bias models under random truncation Authors:  Jacobo de Una-Alvarez - University of Vigo (Spain) [presenting]
Abstract: Random truncation means that the target random variable is observed only when it falls within a random set. This generally results in a sampling bias, in the sense that different values of the target may have different chances to be sampled. We will consider the problem of testing for the null hypothesis of ignorable sampling bias. When the null is true, ordinary estimators are consistent and, therefore, no correction for random truncation is needed. Two different test statistics based on the NPMLE of the sampling probability will be introduced, and their distributional convergence under the null will be justified. Bootstrap algorithms to approximate the null distribution of the test will be presented. Applications to specific forms of truncation will be given. Simulation results and real data analyses will be provided. Possible extensions to general selection bias models will be discussed.