核密度估计下的二分递归算法构建及应用——测算特定收入群体规模的非参数方法拓展

Construction and Application of Bipartite Recursive Algorithm Based on Kernel Density Estimation——a New Non-Parametric Method to Measure the Given Income Population Scale

居民收入密度函数的核密度估计具有非连续性,因无法通过积分计算特定收入区间的人口规模,故在核密度估计基础上,构建二分递归算法用以测算特定收入群体规模。使用中国健康和营养调查中的中国农村居民人均纯收入的微观调查数据,对中国农村居民收入分布进行核密度估计,并通过二分递归算法测算中国农村贫困发生率,结果显示:考虑到微观数据源和数据内容上的一些差异,计算得到的农村贫困发生率符合国家统计局公布的变动趋势且数值差异不大。因此,在核密度估计下使用二分递归算法计算特定收入群体规模具有有效性。

Kernel density estimation of residents＂ income distribution is discontinuous. So a given income population scale can＇t be integral calculated Basing on kernel density estimation this paper constructs bipartite recursive algorithm to measure a given income population scale. Using rural residents＇ micro--data in CHNS, this paper estimates the rural residents＂ income kernel density and poverty rates by bipartite recursive algorithm. The measuring results show that considering the data source and data content difference, poverty rates computed in this paper accord with the trend announced by National Bureau of Statistics. So using bipartite recursive algorithm in kernel density estimation to measure a given income population scale has general validity.