Non-parametric method for filling in the missing value for cross-sectional dataset: a validation on the per capita GDP data at county level in China

Abstract

When dealing with the observation with missing values, we used to get them by means of mathematical interpolation. Compared with the traditional methods for parametric interpolation including linear interpolation, spline interpolation, kriging interpolation, etc., which sometimes export so paradoxical results that there are quite a lot of debates on the reliability of rationale and application, the non-parametric methods are becoming more and more popular to interpolate the missing values for the cross sectional dataset. In this paper, a non-parametric method is introduced and its feasibility of filling in missing values of per capita GDP data at county level for China is illustrated and verified. The results indicate that the nonparametric method produces essentially unbiased estimates by using kernel density function based on a sample drawn from all the observations. So it appears that the actual performance of non-parametric model can be quite helpful to fill in the missing values with a large sample of observation and the non-parametric extrapolation methods tested in this empirical study could be applied in other similar studies.