【摘要】 [目的]结合实例探索评价调查员变异的方法,并探讨当存在调查员变异时,如何分析产生调查员变异的原因以及进一步的数据分析。[方法]以2008年四川地震灾区居民的生命质量及其影响因素相关研究为例,采用多水平模型评价灾区居民SF-12量表的MCS和PCS得分测量中是否存在调查员变异,并比较存在调查员变异时,考虑和不考虑调查员变异对结果的影响。[结果]SF-12量表心理方面的测量存在调查员变异,但SF-12量表生理方面的测量不存在调查员变异;将调查员变异在MCS得分中分离后,模型3中汶川与北川居民MCS得分差异无统计学意义,模型3中房屋垮塌的居民MCS得分比没有房屋垮塌的居民低,均比模型4的结果解释合理,且模型3拟合效果更好。[结论]多水平模型适用于评价调查员变异和探索产生调查员变异的原因,为今后改进调查员培训,提高调查质量提供参考。更多还原

【Abstract】 [Objective]To explore the way to assessing the variation of the interviewers based on the data of a survey,analyze the reasons which result in the variation of the interviewers when there is a variation of the interviewers,and to discuss the potential impact on the results of the modeling fit when the variation of the interviewers is ignored.[Methods]Based on the data from the survey on the quality of life of people in Sichuan quake hit area in 2008,we assessed if there was a variation of the interviewers for the MCS(Mental component summary)and PCS(Physical component summary)of SF-12 by multilevel model,and compared the influence of the outcome considering the variation of the interviewers or not when there was the variation of the interviewers.[Results]For the MCS,there was variation between the interviewers.For the PCS,there was no variation between the interviewers.When the variation of the interviewers was distinguished from the whole variation,the difference of MCS of people in Wenchuan and Beichuan was non-significant in model 3,and the MCS of people whose house were damaged was lower than those whose house were not damaged in model 3,and the results in model 3 were more?reasonable than in model 4.And model 3 had a better modeling fit than model 4.[Conclusion]Multilevel model was alternative method to assess the variation of the interviewers.It was useful to improve the quality of the further interviewer training by exploring the reasons which produce the variation of the interviewers.更多还原