The Application Of Quantile Regression Method In The City Fire Number Research

Quantile regression is a statistical method which can depict different levels of theobject. It is of significance to research some random variables which belong toskewness distribution. Especially, its application is widely used in economic, financeand others. This paper primarily uses quantile regression method to explore thecorrelation between the fire number in different levels and meteorological factors.In the first chapter, we mainly introduce the situation and research significance of fire.Then, the development of quantile regression and the research of forest and urban fireare summarized. In the last, it shows that using the quantile regression method to studythe number of city fire is meaningful.In the second chapter, it presents the quantile regression theory knowledge and thesimilarities and differences between quantile regression and traditional linear regressionto us. After this, we simplely recommend the semiparametric quantile regressionmethod. And in the end of this chapter, it describes how to implement the quantileregression and semiparametric quantile regression in the R software.Base on a city’s fire data of1997-2007, the third chapter adopts quantile regressionapproach to empirically investigate the monthly fire number from the average monthlytemperature, monthly mean relative humidity, monthly average wind speed, monthlyaverage rainfall four aspects. At the same time, we also discuss the nonparameters effectof wind velocity on fire number by using semiparametric quantile regression. At the end,quantile regression equation and semiparametric quantile regression equation, when equal to0.8, are used to forecast all February fire number during this eleven yearsrespectively. The results show that quantile regression can preferably explain the tailcharacteristics of the explanatory variable compared with traditional regression, whilesemiparametric quantile regression is relatively stable to deal with the presence ofabnormal points.The fourth chapter summarizes the whole paper.