| As Hedonic method is used more widely, the importance is self-evident for the study on modeling of Hedonic Price Model (HPM). But the classic HPM must satisfy many statistical assumptions,for example, the normality of commodity prices distribution, and the non-colinearity among the characteristic variables etc. Additionally, characteristic variables and solutions to their colinearity problem are commonly chosen according to stepwise regression method and principal component regression etc. In fact, based on stepwise regression method,the characteristic variables filtered may be too much. And principle component regression method tends to weaken the the economy meaning of them. About these mentioned questions, the detailed researches in this article as follows:Chapter one: Introduction. Based on the application prospect of the HPM,the their economics and statistics hypotheses, the problems on selection variables and the model functional forms, the research background, the status quo home and abroad on HPM, GLM and LARS-Lasso on modeling HPM, theoretical basis and object of thispaper are introduced.Chapter two: Classical Hedonic Price Model and its statistical theory frame on modeling. This article analyses the statistical forms of equilibrium hedonic price, linear HPM, half logarithm HPM, double logarithms, and the HPM based on Box-Cox transform, and makes relatively deep analysis and approach to the defects of characteristic variables selection. The estimation methods of LS, ML and GLS of the shadow price are introduced. For the purpose of the modeling HPM, the evaluation criteria from various angles are introduced too. Especially, the reason has combed by which the collinearity between variables.is evaluated from the condition number.Chapter three: Generalized Linear Model and LARS-Lasso method. This paper outlines GLM function form and its assumptions,introduces the LARS-Lasso method and the estimate method of ML based on Lasso punish of GLM. Chapter four. Statistical framework of modeling Generalized Hedonic Price Model. Based on the GLM theory, generalization Hedonic Price Model is presented. The ML function of Lasso punishment by locally quadratic approximation can be is approximated with one of the LS estimation forms, which realizes piecewise linear estimation of LARS-Lasso. This paper gives evaluation criteria, diagnostic methods, economic explanation, expressions of the marginal hedonic price, commodity pricing strategy, and the methods of compiling LHPI, PHPI and FHPI.Chapter five: Simulation and case analyses. Firstly, when The correlation coefficients of variables are 0.99 and 0.95 respectively and the condition number indicates colinearity obvious, the performance of LARS-Lasso and stepwise regression solving the multicollinearity problems among is validated by simulations. Secondly, based on January 2009 to September notebook computer data, the application value of GHPM theoretical framework is proved. During the research, both the establishment and the forecast of the GHPM are compared with LS estimation method on the whole model, stepwise regression method, and LARS-Lasso method by Various model evaluation criteria. And the model is diagnosed intuitively by statistical techniques. Finally, combining the domestic market, this paper analyzes the laptops pricing and the establishment of the notebook computer HPI.Chapter six: Conclusions and prospected. The main conclusions and further research works are given in this paper.Based on the above aspects of research work, this paper mainly innovation points and conclusions as follows:First, this paper systematically combs the statistical form of HPM economic theory and the statistical framework of modeling HPM, Which enrich and perfect the theoretical framework of modeling HPM.Second, in this paper, modeling of the classical HPM may be grouped in the frame of theoretical GLM, which weakens the their statistical assumptions. The approach how to construct the GHPM availably is illustrated. In this way, the new statistical method LARS-Lasso for some measures on the selecting variables and the shadow prices estimation is presented. The expression of shadow prices of GHPM and the compiling LHPI, FHPI and PHPI methods are given. These should be provides theoretical basis for the application of HPM.Third, through random simulation and taking a real case, the LARS-Lasso method is more effective than stepwise regression method, either in overcoming the colinearity of variables or in selecting variables or in estimating parameters of GHPM. Based on LARS-Lasso method and the modeling HPM theory from GLM,this paper establishes GHPM and combines with industry related information to analyse its economics significance and discusses pricing strategies and prepares HPI. These results verified this theory is feasible and practical. |
