分位回归方法及其在直销中的应用.pdfVIP

摘 要 1978 年，Koneker 和Bassett 首次提出了分位回归方法，该方法可以看成是 条件均值模型中最小二乘估计到条件分位函数估计的推广。由于分位回归具有 很多优点,近年来它逐渐成了线性和非线性响应模型综合性的统计分析方法。 分位回归方法虽然应用广泛，但是很少有研究者将该方法用来做统计预测。 而且该方法应用在直销领域几乎没人做过。本文尝试将分位回归方法应用到直 销领域，并用该方法来预测直销公司的利润，最后利用 “十分位分析法”对预 测结果进行评价。 本文所使用的数据是一家美国直销公司的数据。首先，我们利用一般线性 回归方法对该数据进行统计建模，并根据得到的模型对公司的利润进行预测， 接着利用十分位分析法对预测的结果进行打分；其次，我们对分位回归方法得 到的模型进行与第一步类似的操作；最后，将一般线性回归方法和分位回归方 法的打分结果进行比较，发现利用分位回归方法建立的模型特别是高分位点对 应的模型，其预测效果要比一般线性回归方法好很多。 关键词：分位回归；检验函数；直销；十分位分析 1 Abstract Koneker and Bassett (1978) first proposed quantile regression, which is viewed as an extension of classical least squares estimation of conditional mean models to the estimation of an ensemble of models for several conditional quantile functions. Quantile regression has several useful features and therefore is gradually developing into a comprehensive approach to the statistical analysis of linear and nonlinear response models. Quantile regression has wide applications in all kinds of fields but for statistical prediction especially in direct marketing. One of the main interests of the thesis is to predict company s profits in direct marketing with quantile regression, in which the decile analysis is employed to evaluate the predicted results. The data used in this thesis is from an American direct marketing company s. Firstly, we apply classical linear regression to modeling the data, then use the resulting models to predict profits. The decile analysis is used to evaluate the predicted results. Secondly, we apply quantile regression similar to the first step. Finally, comparing the results from classical linear regression with those from quantile regression, we find that quantile regression model performs better than th