The Modeling And Application Of CGSS Data Quality Evaluation Index System

Most available literature of data quality assessment focuses on official government database and internal database of institution or enterprise,while the data quality of unofficial micro-databases are rarely mentioned.But it's undeniable that these databases play an important role in both achademic research and empirical study.Therefore,this paper devoted to studying the quality of CGSS data,trying to make some useful attempts in the research of data quality of micro data sets,so as to promoting the construction of micro databases.The empirical analysis in this paper can be divided into two parts,that is the construction of index system and the determination of optimal weight.The index system is built after comparisons within the existing evaluation index systems accomplished,data balance indicators was creatively designed to fit the application scenarios.The index system includes indicators of consistency,data integrity,completeness of description,timeliness,richness,reliability and data balance.When comes to the part of determine optional weight,this paper uses academic trace in bibliometric to determine the optional weight of the index system from some common weight assignment methods in terms of determining the weight of each indicators.This method not only free from the influence of subjective interference in objective method,but also help avoid scientific dependence of the objective weighting method as academic trace does not participate in the calculation of the index system score.The model results show that the data quality of CGSS2010,CGSS2015,CGSS2005 scores more than 0.6 within all the published CGSS data.Richness,reliability and data balance takes the first 3 weight than other indexes.Quality problems are summarized from both the questionnaire and data side,including inconstant of data recording methods and incompleteness of data adjustment process,inconsistent data,questionnaire logical inversion,options conflicts with questions etc.The last part of this paper contains the statistical analysis towards the revealed problems,aiming to seek out the causes of them.Suggestions are putted forward accordingly.To solve questionnaire logical inversion and options conflicts with questions,Afford should be made to reinforce the pre-survey process thus the range of data adjustment can be narrowed.Solutions like perfection raw data traceability system or add field research AIDS are put forward to improve inconsistent data,inconstant of data recording methods.Additional suggestions are proposed to improve the data quality in general,including improve the data user communication and feedback channels and deeply tap the respondents' acceptance of different issues to improve the using experience and data completion.