Scholarlink

<p>Partial Least Squares Structural Equation Modeling (PLS-SEM) has gained acceptance in social sciences due to its flexibility with small samples, non-normal data, and complex models. Yet, limited guidance exists on combining different types of higher-order constructs in a single framework. This tutorial presents a unified, practice-oriented roadmap for specifying and validating Type I (reflective–reflective) and Type II (reflective–formative) constructs within a single model. The study contributes by demonstrating an enhanced embedded two-stage approach–operationally implemented as a latent variable-based two-stage approach–that yields second-order validity diagnostics by providing a visually-guided protocol to leverage modern software capabilities, thereby bypassing legacy manual score construction; a design protocol for formative convergent validity that operationalizes redundancy analysis through a pre-planned global single item; and a fit perspective that augments Goodness-of-Fit by incorporating the redundancy coefficient to reflect mixed measurement natures. Results show that multi-type hierarchical models can be estimated effectively and practically using this latent variable-based embedded two-stage approach. The article concludes with methodological guidelines for social science researchers to build more robust and valid multi-type PLS-SEMs.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A tutorial for building multi-type PLS-SEM in social research: validating and testing reflective and formative constructs in a single model

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Gürkan Aybek,
  • Hatice Karakaş

摘要

Partial Least Squares Structural Equation Modeling (PLS-SEM) has gained acceptance in social sciences due to its flexibility with small samples, non-normal data, and complex models. Yet, limited guidance exists on combining different types of higher-order constructs in a single framework. This tutorial presents a unified, practice-oriented roadmap for specifying and validating Type I (reflective–reflective) and Type II (reflective–formative) constructs within a single model. The study contributes by demonstrating an enhanced embedded two-stage approach–operationally implemented as a latent variable-based two-stage approach–that yields second-order validity diagnostics by providing a visually-guided protocol to leverage modern software capabilities, thereby bypassing legacy manual score construction; a design protocol for formative convergent validity that operationalizes redundancy analysis through a pre-planned global single item; and a fit perspective that augments Goodness-of-Fit by incorporating the redundancy coefficient to reflect mixed measurement natures. Results show that multi-type hierarchical models can be estimated effectively and practically using this latent variable-based embedded two-stage approach. The article concludes with methodological guidelines for social science researchers to build more robust and valid multi-type PLS-SEMs.