摘要
运用现代多元统计技术-偏最小二乘回归法(PLS)对影响我国卷烟消费需求的若干因素进行了定量分析。结果表明:本研究设立的数学模型对若干自变量(X)的累计信息利用率达到了100%,同时能够解释因变量(Y)的98.0%变异信息,累计交叉统计有效性达到0.958,预测值与原始观测值组成的数据点都在回归图的对角线附近分布,且回归系数达到0.9795,该数学模型的拟合效果甚佳。研究所选定的若干指标中以第三产业所占GDP比例、人口、卷烟零售价格、商品零售价格基定指数对卷烟消费量的影响最大,其它指标对卷烟消费需求也有明显的影响。
Quantity analyses of factors affecting the demand of cigarettes in China were carried out by modem statistical tech- nology-Partial Least-Squares (PLS) regression. Results showed that the cumulative fraction of the sum of squares (SS) of all the X and Y variables explained by all extracted components were 100% and 98.0% while the cumulative overall cross-validated R2Y for the component was 0. 958. All the points consisted of the observed and predicted values fell close to the 45 degree line of regression plot and the regression coefficient was 0.9795. So, the forecast pattern of per capita cigarette consumption was a good model. Indexes like rate of tertiary industry in GDP, population, cigarette average price and retail price index were most important to the cigarette consumption. Other indexes also had obvious effects on cigarette consumptions.
出处
《中国烟草学报》
EI
CAS
CSCD
2008年第4期44-48,共5页
Acta Tabacaria Sinica
基金
国家发展和改革委员会研究项目"2001-2006间中国卷烟价格变化趋势与消费需求变化分析"(LY0601)
关键词
偏最小二乘回归
卷烟消费
影响因素
partial least squares regression
cigarette consumption
impact factor