| High-quality statistical data is the basis of research,to ensure the quality of statistical data,and to further affirm academic research results.Especially in recent years,the quality of statistical data has become a hot issue of concern to the government and society.With the continuous improvement of China’s economic development level,the international environment has become increasingly polarized.High-quality statistical data can help governments and their institutions to formulate more reasonable Policy system.Scholars have invested a lot of energy in the evaluation of statistical data quality.Different evaluation methods and strategies have been used in the field of data quality detection.However,most scholars have studied macroeconomic data,and there is less research on the quality of micro-survey data.In order to be able to understand the status of social development more comprehensively,many departments and universities have launched many databases composed of micro-survey data.These databases are widely used in the study of social issues.However,at present,there is no systematic method for the statistical evaluation of the quality of micro survey data.The purpose of this study is to propose a more general method for the quality assessment of micro survey data.Combining related theories,three standards for measuring the quality of micro survey data are proposed: representativeness,completeness,and accuracy.The Mayer index method and the United Nations age-sex index method were used to evaluate the representativeness of micro-survey data;the missing rate and variable missing rate were calculated to evaluate its completeness;the logical evaluation method,Benford’s law,and decision tree model were used to evaluate its accuracy.The CFPS database is taken as an example for empirical analysis.In the empirical analysis using the CFPS database as an example,the Maye index method and the United Nations age-sex index method are used to determine the representativeness of the database.The database may have gender preference issues.Second,the CFPS2016 family economic questionnaire is used for completeness and accuracy.According to the calculation,by calculating the missing rate and variable missing rate of this part of the index,we get that the database of this part is relatively complete,but some indexes that require mandatory answers still have more missing values.Interviewers need to improve their inquiry skills and try to Reduce missing values;the comparative logical evaluation method in the logical evaluation method was used to detect three sets of indicators of the CFPS2016 family economic questionnaire.The three groups of indicators did not have fewer logical cases,but there was still room for improvement.Answering “employee income”,but when answering the “worker sent home amount” situation,the number of illogical cases is the largest.In subsequent surveys,interviewers can make targeted improvements;finally,using Benford’s law and related inspection methods,Select the indicator that is most likely to have data quality problems.After analysis,the indicator is " Total cash and deposits ",using the distance detection method to divide the indicator into two parts: abnormal samples and normal samples.The C5.0 algorithm and CART algorithm in the decision tree were used for modeling and analysis,and the C5.0 algorithm decision tree was found.It can classify and predict well,and can find abnormal individuals through this model for subsequent research.This article provides an idea for evaluating the quality of micro survey data,and uses this idea to evaluate the quality of the CFPS database,further verifying the idea and its operability and universal applicability.However,this idea is still imperfect,and there are shortcomings that need to be improved,but it still provides a new reference for the statistical evaluation of the quality of micro survey data. |
