多维贫困指数测定中权重设定的比较研究

发布时间：2018-01-17 21:09

本文关键词：多维贫困指数测定中权重设定的比较研究　出处：《山西财经大学》2017年硕士论文　论文类型：学位论文

【摘要】：对贫困的研究已经越来越从单一的收入维度向多维考察转变,而在多维贫困指数的测算过程当中,不同指标的权重分配是一个非常重要的问题。不同的权重分配有可能会产生截然不同的测算结果,这就必然会使得最终的测算结果损失了部分说服力。因此,对多维贫困测算中同权重设定的比较是非常重要且有意义的。为了对各种权重设定方法进行比较,本文研究了多维贫困研究中权重设定的五种加权方法,包括全指标等权、维度等权后指标等权、主成分分析、多元对应分析和模糊集方法。基于CHNS2011年的数据先计算出各加权方法下各指标的权重值,再分别对全国12个省市的多维贫困进行测算并排序得到五个排序结果。与此同时本文借鉴了蒙特卡罗模拟的思想进行了随机模拟实验,计算得到大量随机实验后各省市的累计排序均值并以此得到稳定的12个省市的排序结果。通过设定总误差函数为单个省市的绝对离差的加总形式,计算得到各加权方法的排序结果的总误差。总误差越大的排序结果所对饮的加权方法越不推荐使用,反之总误差越小则认定为其是更合理的加权方法。按照以上思路并通过实证研究,本文发现,模糊集方法和多元对应分析所得到的总误差最大,达到了12。而全指标等权和主成分分析的总误差较小,分别为4和6。基于该分析结果,本文认为在多维贫困指数测算时应该尽量避免使用模糊集方法和多元对应分析,应该尽可能的使用全指标等权方法或者主成分分析方法。
[Abstract]:The study of poverty has changed from a single income dimension to a multidimensional investigation, and in the process of measuring the multidimensional poverty index. Weight distribution of different indicators is a very important issue. Different weight distribution may produce different results, which will inevitably make the final calculation results lose some persuasiveness. It is very important and meaningful to compare the same weight setting in multidimensional poverty measurement. In order to compare various weight setting methods, this paper studies five weighting methods of weight setting in multidimensional poverty research. It includes full index equal weight, dimension equal weight index equal weight, principal component analysis, multivariate correspondence analysis and fuzzy set method. Based on the data of CHNS2011 year, the weight value of each index under each weighted method is first calculated. Then the multi-dimensional poverty of 12 provinces and cities in China are measured and sorted to get five ranking results. At the same time, this paper uses the idea of Monte Carlo simulation to carry out random simulation experiment. After a large number of random experiments, the accumulative ranking mean of each province and city is calculated and the results of 12 provinces and cities are obtained. By setting the total error function as the summation form of absolute deviation of a single province and city. The total error of the ranking results of each weighted method is calculated. The greater the total error, the less the weighted method of drinking is recommended. On the contrary, the smaller the total error is, the more reasonable the weighted method is. According to the above thinking and through the empirical study, this paper finds that the fuzzy set method and multivariate correspondence analysis can get the largest total error. The total error of total index iso-weight and principal component analysis is smaller, 4 and 6, respectively. Based on the results of the analysis. This paper holds that the fuzzy set method and multivariate correspondence analysis should be avoided in the calculation of multidimensional poverty index, and the whole index equal weight method or principal component analysis method should be used as far as possible.
【学位授予单位】：山西财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F323.8

【参考文献】