Same model, same data, but different outcomes: Evaluating the impact of method choices in structural equation modeling
kHUB post date: November 2024
Originally published: April 4, 2024 (PDMA JPIM • Vol. 41, Issue 6 • November 2024)
Read time: 70 minutes
Access the Full Article
Scientific research demands robust findings, yet variability in results persists due to researchers' decisions in data analysis. Despite strict adherence to state-of the-art methodological norms, research results can vary when analyzing the same data. This article aims to explore this variability by examining the impact of researchers' analytical decisions when using different approaches to structural equation modeling (SEM), a widely used method in innovation management to estimate cause–effect relationships between constructs and their indicator variables. For this purpose, we invited SEM experts to estimate a model on absorptive capacity's impact on organizational innovation and performance using different SEM estimators. The results show considerable variability in effect sizes and significance levels, depending on the researchers' analytical choices. Our research underscores the necessity of transparent analytical decisions, urging researchers to acknowledge their results' uncertainty, to implement robustness checks, and to document the results from different analytical workflows. Based on our findings, we provide recommendations and guidelines on how to address results variability. Our findings, conclusions, and recommendations aim to enhance research validity and reproducibility in innovation management, providing actionable and valuable insights for improved future research practices that lead to solid practical recommendations.
Practitioner Points
- Managers rely on academic research to make informed decisions in their everyday operations. Media outlets often portray research results as unambiguous without acknowledging their uncertainty, which may trigger erroneous conclusions.
- By showcasing the results' variability when analyzing the same model with the same dataset, our study underlines the ambiguity that comes with any statistical analysis—even under relatively controlled conditions.
- Companies should assume different analytical perspectives when working with data in order to safeguard the robustness of the implications drawn from any analysis.