表述效应对平衡量表内部一致性信度的影响:一项蒙特卡洛研究
详细信息 查看官网全文
- 英文论文题名：A Monte Carlo Study of the Effect of Item Wording on the Internal Consistency of Balanced Scale
- 论文作者：韦嘉 ; 张进辅
- 英文论文作者：Jia Wei ; Jinfu Zhang ; Faculty of psychology ; southwest university of China
- 年：2015
- 作者机构：西南大学心理学部;
- 论文关键词：表述效应 ; 共同方法变异 ; 内部一致性信度 ; 蒙特卡洛模拟研究
- 英文论文关键词：item wording effect ; common method variance ; internal consistency reliability ; Monte Carlo study
- 会议召开时间：2015-10-16
- 会议录名称：第十八届全国心理学学术会议摘要集——心理学与社会发展
- 英文会议录名称：Proceedings of the 18th National Academic Congress of Psychology
- 语种：中文
- 分类号：B841
- 学会代码：ZGXG
- 会议名称：第十八届全国心理学学术会议
- 会议地点：中国天津
- 主办单位：中国心理学会
- 学会名称：中国心理学会
- 页数：2
- 文件大小：796k
- 原文格式：D
- 会议级别：全国
摘要
研究者倾向于在量表中加入反向计分条目来平衡受测者的反应定势或反应风格,但研究发现这种策略可能引入新的共同方法变异,产生表述效应,对测量的信效度造成严重的影响。本研究拟通过双因子模型产生模拟数据,考察表述效应对测量内部一致性信度的影响。研究采用3 x 3 x 4 x 4的研究设计,4个自变量分别为条目目的标结构因子负荷(0.35、0.55、0.75)、方法效应因子负荷(0.10、0.20、0.30)、方法因子间的相关(0、0.2、0.4、0.6)以及样本量(100、200、500、1000)。不同的自变量单元均生成1000个模拟样本。研究的因变量为用单因子模型拟合1000个模拟样本的平均α系数与组合信度,通过效应可加模型进行多因素方差分析。结果发现,α系数在不同水平的目标结构因子负荷、方法效应因子负荷与方法因子间的相关上差异显著:F(2,133)=4954.702,F(2,133)=116.265,F(3,133)=5.621(Ps<0.01);组合信度在不同水平的目标结构因子负荷、方法效应因子负荷与方法因子间的相关上差异显著:F(2,133)=4989.765,F(2,133)=116.155,F(3,133)=5.823(Ps<0.01);但α系数与组合信度在不同的样本量上差异不显著F(3,133)=0.675,P=0.569与F(3,133)=0.224,P=0.880。采用Scheffe法进行事后比较发现,α系数与组合信度随目标结构因子与方法因子的升高而升高,但仅在方法因子间无相关与方法因子间高相关时存在显著性差异,后者大于前者。研究结果表明,表述效应会高估测量的信度,应对其加以统计控制或程序控制。
To counteract the participants' response set or acquiescence, the test makers tend to add reversely worded item into the instruments. Such "balanced" strategy have been proved to introduce extra systematical error, common method variance, in other words the item wording effect, which would potentially threaten the reliability and validity of a certain instrument. A Monte Carlo study was employed to explore the item wording effect on the internal consistency of balanced scale. 3x3x4x4 between subject factorial design was employed in the current study. The independent variables were factor loading of the target factor, factor loadings of method factor, the correlation coefficient between two method factors and the sample size. 1000 simulate data were generated by bi-factor model in different combination of the independent variables. The dependent variables were the mean of the Cronbach's α and the composite reliability based on one factor model. The results showed except the sample size, there were significant difference among the levels with respect to the factor loading of the target factor, factor loadings of method factor, the correlation coefficient between two method factors. For Cronbach's α, F(2,133)=4954.702, F(2,133)=116.265,F(3,133)=5.621,(Ps < 0.01), respectively; for composite reliability, F(2,133)=4989.765, F(2,133)=116.155, F(3,133)=5.823,(Ps<0.01), respectively. These results indicated that the item wording effect would overestimate the internal consistency reliability of a certain instrument and should be controlled by statistic or procedure ways.
To counteract the participants' response set or acquiescence, the test makers tend to add reversely worded item into the instruments. Such "balanced" strategy have been proved to introduce extra systematical error, common method variance, in other words the item wording effect, which would potentially threaten the reliability and validity of a certain instrument. A Monte Carlo study was employed to explore the item wording effect on the internal consistency of balanced scale. 3x3x4x4 between subject factorial design was employed in the current study. The independent variables were factor loading of the target factor, factor loadings of method factor, the correlation coefficient between two method factors and the sample size. 1000 simulate data were generated by bi-factor model in different combination of the independent variables. The dependent variables were the mean of the Cronbach's α and the composite reliability based on one factor model. The results showed except the sample size, there were significant difference among the levels with respect to the factor loading of the target factor, factor loadings of method factor, the correlation coefficient between two method factors. For Cronbach's α, F(2,133)=4954.702, F(2,133)=116.265,F(3,133)=5.621,(Ps < 0.01), respectively; for composite reliability, F(2,133)=4989.765, F(2,133)=116.155, F(3,133)=5.823,(Ps<0.01), respectively. These results indicated that the item wording effect would overestimate the internal consistency reliability of a certain instrument and should be controlled by statistic or procedure ways.
引文