Background: A PhD candidate in Technology Management at an IIT-affiliated institute contacted us in March 2025 for help with his data analysis chapter. His research studied the adoption of Industry 4.0 technologies in Indian manufacturing SMEs using PLS-SEM (Partial Least Squares Structural Equation Modelling). He had collected survey data from 312 respondents across 8 industrial clusters.
The challenge: The candidate had attempted to run PLS-SEM in SmartPLS but encountered convergence problems. His measurement model showed several indicators with outer loadings below 0.60, and his structural model had an endogenous construct with R-squared of 0.08, far too low to support his hypotheses. His supervisor had asked for a complete reanalysis with justification for every modelling decision.
What we did: Dr. Patel reviewed the SmartPLS project file and the original questionnaire. The analysis proceeded as follows: (1) three items with outer loadings below 0.50 were dropped after confirming content validity was preserved; (2) Average Variance Extracted (AVE) was recalculated, all constructs achieved 0.50 or above; (3) discriminant validity was assessed using the HTMT criterion rather than the Fornell-Larcker criterion, as required by current SEM reporting standards; (4) the structural model was re-estimated in Python using the semopy library, allowing more granular bootstrapping (5,000 iterations) for path coefficient significance; (5) the low R-squared construct was examined, two theoretically relevant moderators were identified from the literature and added as interaction terms, raising R-squared to 0.41.
Outcome: The revised analysis chapter received a major revision accepted decision from the PhD viva committee, with no further quantitative concerns raised. The moderating effect of digital infrastructure readiness emerged as the most significant finding of the thesis and was subsequently submitted as a standalone paper to an ABDC-ranked journal.
Identifying details anonymised per our confidentiality policy.