I'm dealing with mediation model with latent factors. Before I conducted CFA on my data and all items loaded significantly to the factors. But then I decided to use PLS-SEM for studying mediation, because I have many variables and when I start mediation analysis due to specification model computer can't compute all the information correctly. So when i entered data and started PLS, i got some of items loaded non-significantly, which surprised me, because in all CFAs they loaded on very high significance level.
Could you explain me, what can cause the problem, and maybe there's a literature to read about this.
Covariance-based CFA and SEM is very different from PLS. Therefore, it is not surprising that results (e.g., loadings) may differ. PLS has many problems, see, for example:
Rönkkö, M., Lee, N., Evermann, J., McIntosh, C., & Antonakis, J. (2023). Marketing or methodology? Exposing the fallacies of PLS with simple demonstrations. European Journal of Marketing, 57(6), 1597-1617.
Rönkkö, M., McIntosh, C. N., Antonakis, J., & Edwards, J. R. (2016). Partial least squares path modeling: Time for some serious second thoughts. Journal of Operations Management, 47, 9-27.
Thank you, Dr. Christian Geiser ! So, perhaps, CFA ans SEM are considered as more reliable methods comparing to PLS?
Dear Kateryna,
Unfortunately, the current literature on PLS-SEM is quite confusing as it often states that PLS-SEM is suitable for reflective (and formative) measurement models. In fact, PLS-SEM is not really suitable for models containing latent variables as it produces inconsistent estimates, i.e., even if your sample size converges to infinity your estimates do not converge (in probability) to the true parameter values, given your model is correctly specified. PLS-SEM creates linear combinations of the observed variables as stand-ins for the latent variable and consequently all the parameter estimates associated with the latent variables suffer from attenuation bias. Hence, I would not use PLS-SEM for models comprising latent variables.
Next to the literature that Christian provided you find more literature on my RG profile...
Best regards,
Florian
- Badges
- Science method
- Similar topics
- Methodology
- Research Methodology
- Qualitative Research