心理学
拉什模型
验证性因素分析
比例(比率)
结构效度
样品(材料)
构造(python库)
评定量表
可靠性(半导体)
考试(生物学)
心理测量学
统计
临床心理学
结构方程建模
发展心理学
计算机科学
数学
功率(物理)
古生物学
化学
物理
程序设计语言
色谱法
生物
量子力学
作者
Jian Li,Ahlam Alghamdi,Hua Li,Andrew Lepp,Jacob Barkley,Han Zhang,Ilker Soyturk
标识
DOI:10.1016/j.chb.2022.107552
摘要
This study examined the multidimension assumption of the often-used 33-item, six-factor Smartphone Addiction Scale (SAS) by employing four confirmatory factor models (i.e., one-factor, first-order six-factor, second-order six-factor, and bifactor). Survey data were collected from 1155 undergraduate students in a US public university. Findings showed that the bifactor model was the best fitting model and SAS is a unidimensional instrument. The composite scores made for the six domain-specific factors, often seen in the literature, were not reliable measures for the construct of smartphone addiction and can result in misleading or even incorrect inferential test results. The most reliable and contributing items identified by the bifactor model were selected to form a short, more efficient version of SAS. A Rasch model was performed to test the psychometric properties of the shortened SAS. The new shortened SAS contains 10 items (SAS-10) and had good reliability, construct validity, and no presence of bias towards students in different gender or academic achievement groups. Additional evidence suggests a 4-category rating scale is enough to capture the construct of smartphone addiction. Finally, SAS-10 correlated to numerous external criterion variables similarly to how the extant literature would predict. SAS-10 is provided in appendix C. • The underlying factor structure of the SAS was thoroughly examined. • Survey data from a sample of 1155 U.S. undergraduate students were analyzed. • The original 33-item SAS is not a multidimensional but unidimensional instrument. • The original 6-category rating scale could be replaced by a 4-category one. • A 10-item shortened SAS scale is presented with robust psychometric properties.
科研通智能强力驱动
Strongly Powered by AbleSci AI