统计与数据科学学院知识库

  随着大数据时代的到来，高维数据之间的建模愈显重要，其应用也越来越广泛，但是容易面临“维数灾难”问题。为了更好地解决该问题，本文构建了潜在阈值时变因子扩展的向量自回归（LT-TVP-FAVAR）模型。该模型通过加入因子的方式将更多信息包含在模型中，其次通过加入潜在阈值减少了模型中的待估参数，从而有效解决了高维数据容易出现的“维数灾难”问题，同时校正了脉冲响应函数，并对模型的预测能力进行了改善。鉴于LT-TVP-FAVAR模型在脉冲响应分析和经济变量预测中的优良表现，本文应用该模型研究了经济政策不确定性对宏观经济的影响，并对我国的GDP增长率进行了短期预测。

  关于经济政策不确定性对产出和物价影响的研究，首先分析经济政策不确定性对产出和物价的影响机制，从而为实证数据的选取提供理论依据。其次基于2001年1月至2022年4月间的高维月度数据，应用本文构建的LT-TVP-FAVAR模型实证研究经济政策不确定对产出和物价的影响效应，宏观经济信息涵盖了投资、消费、产出和进出口等26个变量。实证结果表明，本文构建的LT-TVP-FAVAR模型在分析经济变量之间的影响关系时将更多的信息包含在内，使得模型得出的结论与实际情况更加吻合；脉冲响应图显示，经济政策不确定性在短期内对我国的产出和物价水平都具有明显的时变效应，从长期来看冲击效应会逐渐减弱。在经济不稳定时期，我国的经济增长速度变缓，同时我国的物价水平会出现大幅波动。

  关于我国GDP增长率预测的研究，选取2001年1月至2022年12月的GDP增长率以及与其相关的16个经济变量，构建了LT-TVP-FAVAR模型，将其预测结果与基准模型TVP-VAR模型、TVP-FAVAR模型和LT-TVP-VAR模型作对比。LT-TVP-FAVAR模型相对误差为2.675%，小于LT-TVP-VAR模型的4.848%、TVP-FAVAR模型的12.014%和TVP-VAR模型的14.993%，研究结果表明本文构建的模型在GDP预测方面的效果优于选取的基准模型。

     With the advent of the era of big data, modeling between high-dimensional data becomes more and more important, and its application is more and more extensive, but it is easy to face the problem of "dimensional disaster". In order to solve this problem, we construct a vector autoregressive model with extended time-varying factor of laten threshold. The model includes more information in the model by adding factors, and then reduces the parameters to be estimated in the model by adding laten threshold, thus effectively solving the problem of "dimensional disaster" which is easy to occur in high-dimensional data. At the same time, the impulse response function is corrected, and the prediction ability of the model is improved. In view of the excellent performance of the LT-TVP-FAVAR model in impulse response analysis and economic variable forecasting, this thesis studies the impact of economic policy uncertainty on the macroeconomy, and makes a short-term forecast of China's GDP growth rate.

     As for the research on the impact of economic policy uncertainty on output and price, this thesis first analyzes the impact mechanism of economic policy uncertainty on output and price, so as to provide theoretical basis for the selection of empirical data. Secondly, based on the high-dimensional monthly data from January 2001 to April 2022, the LT-TVP-FAVAR model constructed in this thesis is used to empirically study the impact of economic policy uncertainty on output and price. The macroeconomic information covers 26 variables such as investment, consumption, output and import and export. The empirical results show that the LT-TVP-FAVAR model constructed in this thesis includes more information when analyzing the influence relationship between economic variables, which makes the conclusions drawn by the model more consistent with the actual situation. The impulse response chart shows that economic policy uncertainty has obvious time-varying effects on both output and price level in China in the short term, and the impact effect will gradually weaken in the long term. In the period of economic instability, China's economic growth rate will slow down, while China's price level will fluctuate greatly.

     For the research of China's GDP growth rate forecast, the LT-TVP-FAVAR model is constructed by selecting the GDP growth rate from January 2001 to December 2022 and 16 related economic variables. The prediction results were compared with the benchmark models TVP-VAR model, TVP-FAVAR model and LT-TVP-VAR model. The relative error of the LT-TVP-FAVAR model is 2.675%, which is smaller than the 4.848% of the LT-TVP-VAR model, 12.014% of the TVP-FAVAR model and 14.993% of the TVP-VAR model. The research results suggest that the model constructed in this thesis is better than the selected benchmark model in GDP forecasting.