Variance Estimation for SEM in JMP (PLS-SEM)

[statslearner]

Dear JMP Development Team,

I am writing to suggest a valuable enhancement to the Structural Equation Modeling (SEM) platform in JMP, the inclusion of variance-based estimation technique.

Variance estimation plays a critical role in SEM, particularly in providing more reliable and accurate standard errors and confidence intervals, especially when assumptions such as multivariate normality are not fully met. In practical applications, especially in fields like marketing research, behavioral sciences, and social sciences, data often deviate from ideal conditions, making robust variance estimation crucial for valid inferences.

Currently, Partial Least Squares SEM (PLS-SEM) tools such as SmartPLS and WarpPLS have gained significant traction precisely because they emphasize variance-based approaches. They allow researchers to handle complex models with smaller sample sizes and non-normal data structures.

Adding similar or even more flexible variance estimation options in JMP's SEM platform would position it even more competitively in this space. It would also offer an advantage by allowing users to stay within the powerful and user-friendly JMP environment without needing to turn to external tools.

Enhancing the SEM platform with variance-based estimation will not only strengthen model diagnostics but will also expand JMP's reach among both academic researchers and applied industry professionals who are increasingly looking for robust and flexible SEM solutions.

I greatly appreciate the team's efforts in continuously improving JMP and look forward to seeing further advancements in this area.

Thanks

Chandra

[LauraCS]

Thank you so much for submitting this request, Chandra! I couldn't agree more with the importance of providing approaches in SEM for dealing with small sample sizes and deviations from multivariate normality. Although we don't yet have PLS-SEM, we've made some great progress in this direction! Please look at our "Robust Inference" option when data assumptions don't hold (this helps with accuracy of standard errors and confidence intervals). Also, an upcoming new feature makes small sample SEM estimation more effective, with the additional advantage of relaxing all distributional assumptions for predictors! We're pretty excited about this... stay tunned for version 19 improvements =). 

Of course, we'll consider your request to make sure we can provide the very best tools for our users! The information you provided is very helpful!

Thank you,

~Laura C-S