论文部分内容阅读
目的了解目前公共卫生类期刊抽样方法误用现况,初步探讨导致抽样设计错误的影响因素。方法收集整理4种公共卫生期刊杂志2014年01期至12期的全部学术文献,针对常见抽样设计问题,并利用Epi Data3.1软件建立数据库,录入所收集文献的基本信息及抽样设计错误。采用SPSS 18.0软件进行统计描述与决策树分析,初步探讨导致抽样设计错误的影响因素。结果收集整理文献共1 349篇,经过筛选后得到有效文献260篇。其中抽样设计错误个数的均值为3.40个,错误个数>2的文献有172篇,占66.5%。单因素得到2个有统计学意义的变量:概率抽样类型、杂志类别(P<0.05),其χ2值分别为11.457、5.403。按概率抽样类型分层后,多因素分析得到2个决策树模型,经过10层交叉验证后,模型识别总正确率分别为67.4%、73.7%。结论公共卫生类期刊论文所存在的抽样方法误用问题不容乐观,针对不同类型抽样方法及其误用影响因素应采用不同手段加以改进。
Objective To understand the current status of misuse of sampling methods in public health journals and to discuss the influential factors leading to sampling design errors. Methods To collect and compile all the academic documents of 4 kinds of public health periodicals from January 2014 to December 2014, and to solve the common sampling design problems, the Epi Data3.1 software was used to set up a database to record the basic information and sampling design errors of the collected documents. Using SPSS 18.0 software for statistical description and decision tree analysis, preliminary study of the factors that led to sampling design errors. Results A total of 1 349 articles were collected and collected, and 260 valid documents were obtained after screening. The average number of sampling design errors was 3.40, and the number of errors> 2 was 172, accounting for 66.5%. Two variables with statistical significance were obtained by single factor: probability sampling type and magazine category (P <0.05), and their χ2 values were 11.457 and 5.403 respectively. After stratified by probability sampling type, two decision tree models were obtained by multivariate analysis. After 10-layer cross-validation, the overall accuracy of model recognition was 67.4% and 73.7% respectively. Conclusion The misuse of sampling methods in public health journal articles is not optimistic. Different sampling methods and their influencing factors should be improved by different means.