Flexible methods for latent variable models applied to Health Economics

这个项目旨在开发灵活而强大的方法来估计包含未观察到的变量的模型，即潜在变量。这种模型经常被用来在收集的关于个人的数据中捕捉个人的异质性和时间依赖性，每个人都被观察了几个时间段。潜在变量也可以推断群体成员资格，而这种成员资格无法从数据中获得。预期的方法是贝叶斯方法，基于新的粒子方法，允许用户在模型之间进行选择，并预测未来的观测，即使在复杂的情况下也是如此。这项研究旨在通过改进卫生经济学和相关领域的数据使用来为决策提供信息。

This project aims to develop flexible and powerful methods for estimating models containing variables that are unobserved, that is, latent. Such models are often used to capture individual heterogeneity and time dependence in data collected on individuals, with each individual observed for several time periods. Latent variables can also infer group membership, where such membership is unavailable from the data. The intended methodology is Bayesian and based on new particle methods that allow users to select between models and predict future observations even in complex situations. The research aims to inform decision making through improved use of data in health economics and related fields.