Partial Identification of Probability Distributions

En savoir plus sur le livre

The book delves into the concept of partial identification of population parameters, challenging the traditional binary view of identification. It explores how assumptions about the sampling process can enhance inference, even when parameters are not fully identified. Key topics include prediction with missing data, decomposition of finite mixtures, and treatment response analysis. The author systematically examines the implications of empirical evidence and the effects of imposed assumptions on identification regions, adopting a conservative, nonparametric approach throughout.