本文为瑞典斯德哥尔摩大学（作者：Eva Huselius）的学士论文，共47页。

状态空间模型是一种动态模型，它考虑了描述一个时间序列的不可观测组件是如何随时间发展变化的。这将导致估计更少的参数和更小的规范错误。本研究的目的是从潜在的状态空间模型评估单变量时间序列方法，以预测瑞典的通货膨胀率。对指数平滑模型和ARIMA模型（正则模型和状态空间模型）进行了拟合，并与NIER模型进行了比较，结果表明，状态空间MA(9)在NIER模型中表现最好，且具有较低的规范误差。在一个变化的模式时，原始的ARMA(1,11)模型，无论是否具有12季节特性，通常表现良好，但处于太高的水平。在停滞期，状态空间指数（ETS）模型通过捕捉精确的水平而表现良好。结论是，不同的单变量模型在不同的经济周期中表现良好，但多变量状态空间模型在较长时期内可能会更好。

State space models are dynamic models thattake into account how unobserved components describing a time series developover time. This leads to estimation of fewer parameters and smallerspecification errors. The aim of this study was to evaluate univariate timeseries methods from an underlying state space model to predict the Swedishinflation rate. Exponential smoothing and ARIMA models, both regular and froman underlying state space model were fitted, and forecasts were compared withNIER’s. The result showed that a state space MA(9) performed best in relationto NIER, and had lower specification errors. In times of a varying pattern anoriginal ARMA (1,11) model with and without seasonality of 12 often performedwell but at a too high level. In times of stagnation the state spaceexponential (ETS) models performed well, by capturing the accurate level. Theconclusion was that different univariate models can perform well in differenteconomic cycles, but multivariate state space models would probably be betterfor longer periods.

1. 引言

2. 时间序列数据建模

3. 瑞典通胀数据描述

4. 结果

5. 讨论

更多精彩文章请关注公众号：