Start Date
14-12-2012 12:00 AM
Description
Structural equation models are widely used in IS research as they can accommodate both latent and manifest variables. Recent discussion has focused on the relationships between these latent and manifest variables in structural equation models. Despite repeated attempts to clarify the relationships and provide guidelines on their interpretation, there remains much confusion, as evidenced by the ongoing debate about the use of formative measures or formative items. In this paper, we address the topic from a fresh perspective, by establishing a clear separation of theoretical and statistical models, and by introducing the notions of system inputs and outputs. The paper explores the consequences of these principles for the relationships between latent and manifest variables using the computer self-efficacy (CSE) construct as an illustrative example. The novel ideas in modeling allow a fresh perspective on the CSE construct and a resolution of the ongoing debate about the nature of the construct.
Recommended Citation
Evermann, Joerg and Tate, Mary, "An Ontology of Structural Equation Models with Application to Computer Self-Efficacy" (2012). ICIS 2012 Proceedings. 1.
https://aisel.aisnet.org/icis2012/proceedings/HumanBehavior/1
An Ontology of Structural Equation Models with Application to Computer Self-Efficacy
Structural equation models are widely used in IS research as they can accommodate both latent and manifest variables. Recent discussion has focused on the relationships between these latent and manifest variables in structural equation models. Despite repeated attempts to clarify the relationships and provide guidelines on their interpretation, there remains much confusion, as evidenced by the ongoing debate about the use of formative measures or formative items. In this paper, we address the topic from a fresh perspective, by establishing a clear separation of theoretical and statistical models, and by introducing the notions of system inputs and outputs. The paper explores the consequences of these principles for the relationships between latent and manifest variables using the computer self-efficacy (CSE) construct as an illustrative example. The novel ideas in modeling allow a fresh perspective on the CSE construct and a resolution of the ongoing debate about the nature of the construct.