To echo @billi I also would like mediation analysis capabilities in JMP (or a help page with easy to follow guidelines on using bootstrap analyses to calculate indirect effects). Mediation analysis is available not only in SAS, but also in SPSS using the PROCESS macro, and is a valuable tool for researchers.
Yes, JMP really needs something comparable to the PROCESS Macro for SPSS. I much prefer JMP to SPSS in many, many ways, but JMP is pretty much useless to me without these user-friendly functions to help me with complex analyses/designs. Perhaps hire the PROCESS creator as a consultant to help you make this possible? It would make your software so much more attractive to many of us social scientists!
Agree - someone please write a script that can do bootstrapped CI for indirect effects (must plan for more than one mediators). Then work with Andrew Hayes (who wrote the process macro) to work with you guys to do one for JMP that will cover most of his 84 predetermined models in his book (moderation, mediation, mediated-moderation, moderating-mediation etc.)
An example of a simple mediation model is saved in a script in the “Job Satisfaction.jmp” sample data table as “SEM: Path Analysis no Latent”
We also enable specification of simple mediation models through a shortcut:
JMP 16 uses the Joint Significance test to test the indirect effect. Future versions may have options for obtaining the standard error of the indirect effect with the delta method and bootstrapped standard errors.
It's great to see there's interest in moderated-mediation and mediated-moderation too. Currently, these models are supported in SEM by explicitly creating interaction variables in the data table and using them when launching the platform. We plan to make some of the most common PROCESS macro models available in our "Model Shortcuts" in the future.
Great to see how much progress has been made on this. One lacking feature/issue is that for Model Variables "All columns must have numeric data type and continuous modeling type." In the social sciences, it is typical that a continuous mediator would link a binary IV and/or binary DV (the former being more important). Since the SEM platform can only handle continuous data types, this substantially limits social scientists' use of the platform. Allowing binary variables (similar to the PROCESS Macro for SPSS) would be an excellent next step.
When running analysis, we often use variables that have only two states, but they are nonetheless continuous. One example is dichotomous variables. Another is when using dummy variables to represent categorical variables. Many researchers would define them as continuous, even if they only accept a zero or a one.