Background: Subjective survival probabilities incorporate individuals’ view about own future survival and they are associated with actual mortality patterns. Objective: The objective of this study is twofold. First, we apply a Bayesian methodology to incorporate the respondents’ views about future survival, and second, we investigate whether subjective survival information is useful for predicting actual mortality and self-reported change in health. Methods: To achieve the above-mentioned objective, we adopt a two-step process. In the first step, we use a Bayesian linear regression model, under default priors, on the logit transformation of the subjective mortality probabilities to estimate the posterior distribution of the regression coefficients of the available explanatory variables. In the second step, we fit Bayesian logistic regression models on actual mortality and self-reported change in health, using a variety of priors derived from the posterior distributions of the first step Bayesian model. Data from the Health and Retirement Study (HRS) Waves 13 and 14 are used in this paper. Results: We conclude that the additional information incorporated via the subjective mortality probabilities is useful for predicting actual mortality but less useful for predicting selfreported change in health. Contribution: The contribution of this study relates to the development of a procedure, which can be used to include prior information – based on subjective survival views – in hierarchical Bayesian regression models to improve the ability to predict mortality and self-reported change in health.

中文翻译：

---

将主观生存信息纳入死亡率和健康状况预测变化：贝叶斯方法（作者：Apostolos Papachristos、Dimitrios Fouskakis）


背景：主观生存概率包含了个人对自己未来生存的看法，并且与实际死亡率模式相关。目的：本研究的目的是双重的。首先，我们应用贝叶斯方法来纳入受访者对未来生存的看法，其次，我们调查主观生存信息是否有助于预测实际死亡率和自我报告的健康变化。方法：为了实现上述目标，我们采用两步过程。第一步，我们在默认先验条件下使用贝叶斯线性回归模型对主观死亡率概率进行 Logit 变换，以估计可用解释变量的回归系数的后验分布。在第二步中，我们使用从第一步贝叶斯模型的后验分布导出的各种先验，对实际死亡率和自我报告的健康变化进行贝叶斯逻辑回归模型的拟合。本文使用健康与退休研究 (HRS) 第 13 波和第 14 波的数据。结果：我们的结论是，通过主观死亡率概率纳入的附加信息对于预测实际死亡率很有用，但对于预测自我报告的健康变化不太有用。贡献：这项研究的贡献涉及一种程序的开发，该程序可用于将基于主观生存观点的先验信息纳入分层贝叶斯回归模型中，以提高预测死亡率和自我报告的健康变化的能力。