Examining Differences in Within- and Between-Person Simple Structures of an Engineering Qualification Test Using Multilevel MIMIC Structural Equation Modeling

Frontiers in Applied Mathematics and Statistics

Sideridis, Georgios D.¹  Tsaousis, Ioannis²  Al-Harbi, Khaleel⁴ 

[1] Boston Children'Department of Psychology, University of Crete, Greece;National and Kapodistrian University of Athens, Greece;s Hospital, Harvard Medical School, United States

关键词: Multilevel Structural Equation Modeling (MLSEM);    Nested models;    Multilevel reliability;    construct validity;    Multiple Indicator Multiple Causes Model (MIMIC);   

DOI  :  10.3389/fams.2018.00003

学科分类：数学（综合）

来源: Frontiers

【 摘 要 】

The current study sought to meet three aims: (a) to understand the optimal factor structure of the Professional Engineering (ProfEng) test, a measure aiming to assess competency in engineering, within a multilevel (nested) perspective; (b) to examine the psychometric measurement invariance of the ProfEng test across levels due to nesting and across gender at the person level, and, c) to examine the internal consistency of the engineering competency measure at both levels in the analysis. Data involved 1,696 individuals across 21 universities who took a national licensure test as part of the professional accreditation process to obtain a work permit and practice the engineering profession in the Kingdom of Saudi Arabia. Data were analyzed by use of Multilevel Structural Equation Modeling (MLSEM). Results indicated that a 2-factor model at both levels of analysis provided the best fit to the data. We also examined violation of measurement invariance across clusters (cluster bias). Results showed that all factor loadings were invariant across levels, suggesting the presence of strong measurement invariance. Last, invariance across gender was tested by use of the MIMIC multilevel model. Results pointed to the existence of significant differences between genders on levels of personal and professional skills with females having higher levels on personal skills and males on professional. Estimates of internal consistency reliability also varied markedly due to nesting. It is concluded that ignoring a multilevel structure is associated with errors and inaccuracies in the measurement of person abilities as both measurement wise and precision wise the multilevel model provides increased accuracy at each level in the analysis.

【 授权许可】

CC BY