BUSINESS ADMINISTRATION RESEARCH FORMATIVE INDICATOR MEASUREMENT: HOW TO DEAL WITH THEIR MULTICOLLINEARITY

The use of formative indicators in structural equation models is one of the reasons for using the Partial Least Squares Path Modeling (PLS-PM) instead of LISREL. Furthermore, the use of PLS-PM in Business Administration researches has increased. The objective of this study is toevaluate the effects...

Descripción completa

Detalles Bibliográficos
Autores: Bido, Diógenes de Souza, Silva, Dirceu da, Souza, Cesar Alexandre de, Godoy, Arilda Schmidt
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2010
País:Brasil
Institución:Associação Nacional dos Cursos de Graduação em Administração (ANGRAD)
Repositorio:Administração (São Paulo. Online)
Idioma:portugués
OAI Identifier:oai:ojs.raep.emnuvens.com.br:article/145
Acceso en línea:https://raep.emnuvens.com.br/raep/article/view/145
Access Level:acceso abierto
Palabra clave:structural equation modeling
SEM
LISREL
PLS-PM
formative indicators.
modelagem em equações estruturais
MEE
indicadores formativos
Descripción
Sumario:The use of formative indicators in structural equation models is one of the reasons for using the Partial Least Squares Path Modeling (PLS-PM) instead of LISREL. Furthermore, the use of PLS-PM in Business Administration researches has increased. The objective of this study is toevaluate the effects of the formative indicators' multicollinearity on the estimated values for the factor weights and the structural coefficient. To simulate different levels of multicollinearity, the number of indicators per latent variable and the correlation between them are varied, sogenerating 540 models with different values for the structural coefficients. As conclusion, it is found that the growth of factor weight variability is increased for lower values of structural coefficient. Despite the impossibility of evaluating the relative importance of each indicator tomeasure the construct under the influence of multicollinearity, it is observed that the structural coefficients are not altered. There is also observed inconsistency of PLS-PM when using less than five indicators per latent variable (consistency at large) and when the reliability is less than 0.9. In the end, some recommendations are made so as to minimize the effects of multicollinearity and some directions are given towards further researches.