Mover-Stayer Models for the Analysis of Event-History Data in Social Science Research: A Project Description

本项目开发了三组用于分析事件历史数据的移动-停留模型。移动-停留模型采用一对回归：“s回归”表示停留者的比例，其中停留者是发生事件风险为零的潜在“人”群体；“m回归”表示在移动者中发生事件的条件率，其中移动者是发生事件风险非零的潜在人群。动态模型提供的信息是传统危险率模型所不能提供的。本研究通过以下方式进一步发展了移动-停留模型：(1)扩展了一般的移动-停留模型，允许使用时变预测因子；(2)开发了专门用于分析扩散过程（如信息、疾病或创新的扩散）的移动-停留模型，以确定更高扩散率的预测因子是否意味着更快速的扩散，还是受扩散事件影响的人数更多，或者两者兼而有之；(3)建立具有缺失信息的事件历史数据的推动者-滞留者模型。将这些新的方法工具应用于分析第一次婚姻和再婚、药物使用进展和住宅流动，将提供重要的新的实质性见解。

This project develops three groups of mover-stayer models for the analysis of event-history data. Mover-stayer models employ a pair of regressions: an `S-regression` for the proportion of stayers, where stayers are a latent population of `people` with zero risk of having the event, and an `M-regression` for the conditional rate of having the event among movers, where movers are a latent population of people with nonzero risk of having the event. Mover-stayer models provide information that cannot be obtained from the results of conventional hazard rate models. This research further develops mover-stayer models by: (1) extending general mover-stayer models to permit the use of time-varying predictors (2) developing mover-stayer models specifically designed for the analysis of diffusion processes (such as diffusion of information, diseases, or innovation) to identify whether a predictor of a higher rate of diffusion implies either more rapid diffusion, or a larger number of people who are affected by the diffusion event, or both; (3) developing mover-stayer models for event-history data with some missing information. Applications of these new methodological tools to analyses of first marriages and remarriages, drug-use progression, and residential mobility will provide important new substantive insights.