分位数回归在房地产行业的应用

发布时间：2018-07-12 16:29

本文选题：分位数回归 + 时间序列分析　；参考：《温州大学》2013年硕士论文

【摘要】：分位数回归是Koenker和Bassett针对最小二乘的不足提出来的一种新的估计方法，它不仅能够度量回归变量对分布中心的影响，而且能度量回归变量对分布上尾和下尾的影响，在不同的分位数下进行预测，得到的信息更为全面和精确。时间序列分析是对时间序列数据建立模型,分析数据的内部结构和自身规律,从而对未来的发展进行预测。用分位数回归方法来估计时间序列模型时,对随机误差的分布不做要求,能更加全面的刻画分布的特征。本论文首先介绍了分位数回归的基本理论和性质。其次应用线性分位数回归理论，并结合多元统计的相关知识，给出了房地产行业发展影响因素的模型，从定量角度把握各指标之间的数量关系，得出在房价不同的地区，经济，，环境，房地产自身和人口因素对房价的影响程度也不同的结论。第三详细介绍了ARMA模型的相关理论与建模过程，为ARMA模型在后面的实际应用中提供了理论基础。最后结合分位数回归与ARMA模型对国房景气指数进行实证分析，得到不同分位数下的AR (1)模型，从而可以针对不同水平的数据采用不同的模型进行序列的模拟和预测，文章还通过对比预测值和真实值的差异，明确指出相比最小二乘估计，用分位数回归估计出的AR (1)模型更为准确。
[Abstract]:Quantile regression is a new estimation method proposed by Koenker and Bassett for the deficiency of least squares. It can not only measure the influence of regression variables on the distribution center, but also measure the influence of regression variables on the upper tail and lower tail of distribution. The information obtained is more comprehensive and accurate when the prediction is carried out under different quantiles. Time series analysis is to build a model of time series data, analyze the internal structure of the data and their own laws, so as to predict the future development. When the quantile regression method is used to estimate the time series model, the distribution of random errors is not required, and the characteristics of the distribution can be described more comprehensively. In this paper, we first introduce the basic theory and properties of quantile regression. Secondly, by applying the linear quantile regression theory and combining the related knowledge of multivariate statistics, the paper gives the model of the influencing factors of the development of the real estate industry, from the quantitative angle to grasp the quantitative relationship between the various indicators, and draws the conclusion that in the regions with different housing prices, the economy. The influence of environment, real estate and population on house price is different. The third part introduces the theory and modeling process of ARMA model in detail, which provides a theoretical basis for the practical application of ARMA model in the future. Finally, combining the quantile regression and ARMA model, we get the AR (1) model under different quantiles, so we can use different models to simulate and predict the different levels of data. By comparing the difference between the predicted value and the real value, it is clearly pointed out that the AR (1) model estimated by quantile regression is more accurate than the least square estimation.
【学位授予单位】：温州大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F299.23;F224;O212.1

【参考文献】