Percentile Regression Methods In Financial Management

For management science researcher who conducts the quantitative analysis research, the ordinary least squares (OLS) expression has already become a familiar, reliable tool for the analysis of the large-scale and empirical problem(Western, 1995). But when we use the least squares method, the condition is quite high, such as the linear regression model requests to satisfy the homoscedasticity, the random error between two is non-correlated and so on. When we need to carry on the significance inference of the regression coefficient, we usually must suppose the residual error obeys normal distribution. In practical problems, the situations which fully meet these basic assumptions are rare, but once a basic assumption is violated in the application, then it will be difficult to obtain unbiased and effective parameter estimator.This paper is to resolve the problem, take mixed cost's decomposition and OILPLUS company fuel consumption distribution as the main object of study, when carring on the parameter estimation, I propose the quantile regression method, and compare with the ordinary least squares method.The main contents are:The first chapter is the introduction, elaborate the background of quantile regression method; summarizes the domestic and foreign study situation; article frame.The chapter 2 is the theoretical foundation, introduced the concept of quantile; quantile regression model; the nature of quantile estimator that is consistency and gradual normality.The chapter 3 apply quantile regression method to financial management, simulate mixed cost's decomposition and OILPLUS company fuel consumption distribution, analysis the difference between the least squares regression method and quantile regression method, obtain more rich information compared with the least squares.The chapter 4 is the conclusion and the prediction, summary the full text; point out the study deficiency.