Abstract

Hierarchical component modelling (HCM) has become an indispensable analytical strategy in information systems (IS) research for representing multi-componential phenomena, such as technology readiness, organisational capability, user experience, and digital transformation. Despite the substantial methodological literature, recent IS publications reveal persistent inconsistencies in the specification, estimation, and reporting of higher-order constructs. These inconsistencies are particularly pronounced among early-career researchers, who often lack integrated, decision-oriented training in advanced PLS-SEM modelling, an asymmetry that is especially acute for scholars from emerging economies. This TREO Talk introduces a pedagogically structured tutorial that consolidates dispersed methodological guidance into a single replicable workflow. The template guides researchers through five sequential stages: (i) theoretical justification for hierarchical modelling; (ii) specification of the four HCM types (reflective–reflective, reflective–formative, formative–reflective, and formative–formative) with the corresponding estimation approaches (repeated indicators, two-stage, and embedded two-stage); (iii) lower-order construct measurement-model assessment; (iv) higher-order construct measurement and structural-model evaluation, including bootstrapped path inference, R², f², and PLSpredict-based predictive relevance; and (v) transparent reporting and robustness diagnostics aligned with leading IS journal expectations. To anchor the methodology in concrete practice, the tutorial draws on a study of Generative AI (GenAI) readiness in developing-country higher education institutions. The example demonstrates how an originally disaggregated specification - encumbered by over ten structural relationships and discriminant validity concerns (HTMT 0.70–0.85; VIF > 3.0) - was reconceptualised as a parsimonious model with two formative higher-order constructs (Technical Infrastructure, Digital Competency), one reflective higher-order construct (GenAI Readiness), a first-order mediator (Technical Feasibility), and a formative moderator (Institutional Context). The findings indicate that the embedded two-stage approach suits formative higher-order constructs entering interaction terms and that conflating outer weights with outer loadings remains a recurrent reporting error. The authors seek to share the tutorial and decision framework with the IS community (particularly doctoral candidates and early-career faculty in the Global South) to invite critique that may refine its sequencing, worked example, and reporting protocol prior to formal publication. Feedback on target outlets and cross-institutional capacity-building is especially welcomed.

Download

Share

COinS