For many applications of agent-based models (ABMs), an agent's age influences important decisions (e.g. their contribution to/withdrawal from pension funds, their level of risk aversion in decision-making, etc.) and outcomes in their life cycle (e.g. their susceptibility to disease). These considerations make it crucial to accurately capture the age distribution of the population being considered. Often, empirical survival probabilities cannot be used in ABMs to generate the observed age structure due to discrepancies between samples or models (between the ABM and the survival statistical model used to produce empirical rates). In these cases, imputing empirical survival probabilities will not generate the observed age structure of the population, and assumptions such as exogenous agent inflows are necessary (but not necessarily empirically valid). In this paper, we propose a method that allows for the preservation of agent age-structure without the exogenous influx of agents, even when only a subset of the population is being modelled. We demonstrate the flexibility and accuracy of our methodology by performing simulations of several real-world age distributions. This method is a useful tool for those developing ABMs across a broad range of applications.
翻译:对于许多代理模型应用程序而言,代理年龄会影响重要决策(例如他们对退休金的贡献/取款、在决策制定中的风险规避水平等)和生命周期中的结果(例如他们易感性)。这些考虑因素使得准确捕捉所考虑人群的年龄分布至关重要。由于样本或模型(代理模型和用于生成经验速率的生存统计模型之间)之间存在差异,因此经验生存概率通常不能用于在代理模型中生成观察到的年龄结构。在这些情况下,使用经验生存概率进行插补不会生成人口的观察到的年龄结构,因此必须采用外源性代理流入等假设(但未必符合经验)。在本文中,我们提出了一种方法,可以在不插入外源流的情况下保留代理年龄结构,即使只模拟了人口的子集。我们通过多个真实世界年龄分布的模拟展示了我们的方法的灵活性和准确性。该方法是跨各种应用程序开发代理模型的有用工具。