Multi-level Risk-interdependence Models for Competing Events and their Applications to Social and Demographic Research

该项目开发了经改进的危险率模型，用于分析两种或两种以上生活事件的发生情况，如（1）不同原因造成的死亡，（2）婚姻和同居，（3）已婚生育和就业妇女自愿离职，或（4）开始使用和重新使用风险并非独立存在的不同类型的非法药物。 这些模型区分两种类型的风险相互依赖性，通过扩展现有的模型分析这种竞争事件。 一种是由于危险的直接相互依赖性，例如风险相互依赖性，当特定死亡原因的风险增加时会增加其他死亡原因的风险，反之亦然。 另一种是由于未观察到的个体差异，如风险相互依赖性，当一个人有某种无法测量的发病率时，他就有更高的多种死亡原因的风险。 被称为相关脆弱性模型的竞争事件分析的现有模型主要考虑后一种而不是前一种类型的风险相互依赖性，而当前项目将开发模型来解释这两种类型。 该项目还将开发"因果模型"，用于分析一个事件对另一个事件发生的影响，如婚前同居对婚姻的影响，或吸食大麻对吸食其他非法药物的影响。 两个生活事件之间的风险相互依赖性产生了经历第一个事件的人对第二个事件风险的非随机性。 新模型考虑了风险相互依存的两个方面，消除了在选择第一个事件的经历者时存在的偏差，从而可以评估经历第一个事件对第二个事件发生的因果影响，从而有助于更深入地理解生活事件之间的相互依存关系。 它还将提供更好的因果分析，说明生活事件对另一事件发生的可能性的可能影响。 特别是，为了研究工作场所的家庭友好度的差异如何影响这两种生活事件之间的关系，本项目将分析日本就业女性的婚内生育和自愿离职之间的相互依存关系。

The project develops improved hazard rate models for analysis of the occurrence of two or more life events, such as (1) death by different causes, (2) marriage and cohabitation, (3) marital childbirth and voluntary job separation among employed women, or (4) initiation and relapse of different types of illicit drugs whose risks are not independent. These models distinguish two types of risk interdependence by extending existing models for the analysis of such competing events. One type is due to a direct interdependence of hazards such as risk interdependence which exists when an increase in the risk of a particular cause of death increases the risk of other causes of death, and vice versa. The other is due to unobserved individual differences such as risk interdependence which exists when a person who has a certain unmeasured morbidity is at higher risk of multiple causes of death. Existing models for analysis of competing events known as correlated frailty models consider mainly the latter and not the former type of risk interdependence, while the current project will develop models to account for both types. This project will also develop "causal models" for analysis of the influence of an event on the occurrence of another event, such as the influence of premarital cohabitation on marriage, or the influence of initiation of marijuana on initiation of other illicit drugs. The risk interdependence between two life events generates non-randomness of people who experience the first event regarding the risk of the second event. By taking into account the two aspects of risk interdependence, the new models can assess the causal effect of experiencing the first event on the occurrence of the second event by eliminating bias in the selection of people into the experience of the first event.This research will thus contribute to a deeper understanding of interdependence among life events. It will also provide better causal analyses regarding the possible influence of a life event on the likelihood of occurrence of another event. In particular, the project will analyze the interdependence between marital childbirth and voluntary job separation among employed women in Japan in order to study how differences in family-friendliness of the workplace affect the relationship between these two life events.