Binary response models comparison using the 𝛼-Chernoff divergence measure and exponential integral functions

In this paper, the families of binary response models are describing the data on a response variable having two possible outcomes and p p explanatory variables when the possible responses and their probabilities are functions of the explanatory variables. The α \alpha -Chernoff divergence measure and the Bhattacharyya divergence measure when α = 1 / 2 \alpha = 1/2 are the criterion functions used for quantifying the dissimilarity between probability distributions by expressing the divergence measures in terms of the exponential integral functions. The dependences of odds ratio and hazard function on the explanatory variables are also a part of the modeling.

Fragility of identification in panel binary response models

Summary The present paper considers a linear binary response model for panel data with random effects that differ across individuals but are constant over time, and it investigates the roles of the various assumptions that are used to establish conditions for identification. The paper also shows that even for this simple model, it is always possible—including in the logistic case—to find a distribution of the random effects given the exogenous variables, such that the slopes' parameters are arbitrarily different, but the joint distributions of the binary response variables are arbitrarily close.

SMOOTHED QUANTILE REGRESSION PROCESSES FOR BINARY RESPONSE MODELS

In this article, we consider binary response models with linear quantile restrictions. Considerably generalizing previous research on this topic, our analysis focuses on an infinite collection of quantile estimators. We derive a uniform linearization for the properly standardized empirical quantile process and discover some surprising differences with the setting of continuously observed responses. Moreover, we show that considering quantile processes provides an effective way of estimating binary choice probabilities without restrictive assumptions on the form of the link function, heteroskedasticity, or the need for high dimensional nonparametric smoothing necessary for approaches available so far. A uniform linear representation and results on asymptotic normality are provided, and the connection to rearrangements is discussed.