The parametric estimators applied by rolling are commonly used in the analysis of time series with nonlinear features, such as structural change due to time varying parameters and local trends. This paper examines the properties of rolling estimators in the class of Temporally Local Maximum Likelihood (TLML) estimators. We study the TLML estimators of constant parameters, stochastic and stationary parameters and parameters with the Ultra Long Run (ULR) dynamics bridging the gap between the constant and stochastic parameters. Moreover, we explore the properties of TLML estimators in an application to the Susceptible-Infected-Susceptible (SIS) epidemiological model and illustrate their finite sample performance in a simulation study.
翻译:在分析具有非线性特征的时间序列时序时,通常使用滚动的参数估计值,如因时间和当地趋势而发生的结构变化;本文件审查TTLML(TLML)最大局部隐性估计值类别中滚动估计值的特性;我们研究TLML对恒定参数、随机和固定参数和参数的估计值,以及超长运行动态,以弥合恒定参数与随机参数之间的差距;此外,我们探讨TLML估计值在应用可感知、可感知、可视(SIS)流行病学模型时的特性,并在模拟研究中说明其有限的样本性能。