心理科学

心理科学 ›› 2011, Vol. 34 ›› Issue (5): 1195-1200.

• 统计与测量 • 上一篇    下一篇

二阶因素模型的测量等价性检验——以大学生网络利他行为量表为例

郑显亮1,顾海根2,赵必华3   

  1. 1. 赣南师范学院
    2. 上海师范大学
    3. 安徽师范大学
  • 收稿日期:2011-04-18 修回日期:2011-06-14 出版日期:2011-09-20 发布日期:2011-09-28
  • 通讯作者: 顾海根
  • 基金资助:

    上海市重点学科“发展与教育心理学”资助

Testing Measurement Invariance of Second-Order Factor Models: taking Internet Altruistic Behavior Scale of Undergraduates as an example

Xian-Liang ZHENG,Gu Haigen,Zhao Bi-Hua   

  • Received:2011-04-18 Revised:2011-06-14 Online:2011-09-20 Published:2011-09-28
  • Contact: Gu Haigen

PDF

摘要：

与一阶因素模型相比，二阶因素模型具有较多优点，但二阶因素模型的测量等价性检验要更复杂，它需要依次进行七个不同水平的检验：形等价、一阶弱等价、二阶弱等价、一阶强等价、二阶强等价、二阶严等价和一阶严等价。低水平的等价性满足之后，才能进行更为严格的高一水平的等价性检验。运用均值和协方差结构（MACS）模型对大学生网络利他行为量表（IABSU）进行二阶因素模型的测量等价性检验，结果表明，IABSU具有跨地域的完全一阶、二阶严等价性。

Abstract:

Before investigating any potential between-group differences, it is necessary to test the measurement invariance between groups. Some researchers discussed the measurement invariance of first-order factor model, but the measurement invariance research of second-order factor model was rare. Compared with first-order factor model, the second-order factor model has several advantages, but testing measurement invariance of second-order factor model was more complex. In the paper, the measurement invariance across regions in the Internet Altruistic Behavior Scale of Undergraduates (IABSU) was tested. IABSU which include 26 items comprised four first-order factors and single second-order factor. Items were rated on a 4-point Likert-type scale ranging from 1 (“never do”) to 4 (often do). A total of 592 college students from five universities in Shanghai City and Anhui Province were examined using random sampling method. The 283 college students from Shanghai were taken as the developed region sample, the mean age was 20.1 (SD=1.2). And the 309 college students from Anhui Province were taken as the developing region sample, the mean age was 20.4 (SD=1.1). First, baseline model was established to fit the data of the developed region and the developing region separately. The results showed that the absolute index RMSEA and SRMR were less than .08, the relative index NNFI and CFI were more than .95, which indicated that the model could extremely fit the data. And then, the measurement invariance of second-order factor model of IABSU was tested, following the general procedures suggested by Widaman and Reise (1997), including configural invariance, first-order weak invariance, second- order weak invariance, first-order strong invariance, second-order strong invariance, second-order strict invariance and first-order strict invariance. After the lower level invariance was achieved, the higher level invariance which was more stringent could be tested. All procedures were based on the analysis of mean and covariance structures (MACS) within the framework of CFA modeling. The results showed that IABSU was full first-order and second-order strict invariance across regions. In the paper, we showed clearly how to test the measurement invariance of second-order factor model taking IABSU as an example, and also indicated the measurement invariance across regions of IABSU was good, which could be used to test the internet altruistic behavior in different regions.