MULTITRAIT-MULTIMETHOD INFORMATION MANAGEMENT: GLOBAL STRATEGIC ANALYSIS ISSUES
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
International researchers commonly encounter fitting problems when analyzing Multitrait- Multimethod data with confirmatoryfactor analysis (CFA) in both domestic and cross-cultural research (Scherpenzeel and Saris I997). In 1992, Kenny and Kashy examined a special case ofone ofthe mostfrequently used CFA models using one data set and found that it was not statistically identified and, therefore, not estimable. They then argued that most MTMM data were likely to con;form closely to this special case, and because of this, alternative CFA models should be utilized in most data analysis. In this study nine data sets are analyzed using five different additive factor analytic models (CFA with inequality constraints, CFA with error variances.fixed to uniqueness estimates from exploratory factor analysis, Rindskopf parameterization, correlated errors model, and a combination o f the Rindskopf corre- lated errors model) and one multiplicative factor analytic (MUTMUM) model. As with previous attempts, myriad fitting problems were encountered with all six alternative specifications. Closer inspection indicates fitting prob- lems may be due to imperfections in the data rather than fundamental flaws with the model specifications - indicat- ing possible probb!ms with the robustness ofthe techniques utilized. Several suggestionsfor improving international research to possibly avoid these problems are provided as are alternative methods ofestimating construct validity.
Abstract
International researchers commonly encounter fitting problems when analyzing Multitrait- Multimethod data with confirmatoryfactor analysis (CFA) in both domestic and cross-cultural research (Scherpenzeel and Saris I997). In 1992, Kenny and Kashy examined a special case ofone ofthe mostfrequently used CFA models using one data set and found that it was not statistically identified and, therefore, not estimable. They then argued that most MTMM data were likely to con;form closely to this special case, and because of this, alternative CFA models should be utilized in most data analysis. In this study nine data sets are analyzed using five different additive factor analytic models (CFA with inequality constraints, CFA with error variances.fixed to uniqueness estimates from exploratory factor analysis, Rindskopf parameterization, correlated errors model, and a combination o f the Rindskopf corre- lated errors model) and one multiplicative factor analytic (MUTMUM) model. As with previous attempts, myriad fitting problems were encountered with all six alternative specifications. Closer inspection indicates fitting prob- lems may be due to imperfections in the data rather than fundamental flaws with the model specifications - indicat- ing possible probb!ms with the robustness ofthe techniques utilized. Several suggestionsfor improving international research to possibly avoid these problems are provided as are alternative methods ofestimating construct validity.