For those of you that have experience using Salford Systems CART, what do you like better about the JMP Partition platform? I've always found JMP to be a fantastic and poweful tool but have many colleagues that use Salford instead. Are there certain cases where JMP is a much better system for decision tree analytics? Will not having JMP Pro limit my ability to really utilize the data mining capabilities available in JMP?
I can only give an opinion, not an answer, to your last question...having zero experience with the CART method you mention.
Not having JMP Pro will not limit your ability to utilize the modeling methods available in JMP. The fact that you don't have JMP Pro does not influence the JMP user's ability to use any of the capabilties in JMP.
Maybe the question you were really asking is; 'Will not having JMP Pro limit my ability to solve the practical problems I may have?'...absolutely...if for no other reason the tree based (and other advanced modeling techniques such as, penalized regression methods, cross validation, model comparison, and scoring code generation) modeling methods alone are much richer in JMP Pro compared to JMP.
I was an early user of Salford Systems' CART software. It had great "street cred" and capabilities, but the user interface made we want to repeatedly punch my monitor. It was monumentally awful getting data into the software and worse getting information out. Preparing presentations with CART required making static screenshots, and physically transcribing the information into another analysis tool. Customer service and support was abyssmal. I put up with the bad UI and support because the capabilities of the CART software was so good.
When JMP started it's Partition platform, it lacked several important features that were available in CART. The Partition platform has been improved over several versions, and it is really great now. Salford Systems' CART may have improved over time too, but I no longer need to put up with the negatives of their UI and support...JMP's Partition platform does everything I need now. I think JMP's Partition platform is the better system. It's easy to get data into JMP and get information and pretty presentations out. And I also punch many fewer monitors. :)
You should get JMP Pro.
"Will not having JMP Pro limit my ability to really utilize the data mining capabilities available in JMP? "
I've been able to get a lot done in JMP, but when it comes to true data mining problems, the short answer is an emphatic YES!
Head to head CART and Partition? Since Partition is integrated with all the other tools in JMP, getting to root cause is faster, easier and documentable. Partition has a lot of flexibility to find best splits or manualy drive the process, it even lets out lock out a variable and then unlock and keep going.
When you move in to bootstrap forest, and boosted trees, you're also moving into an arena where sampling, overfitting, tuning and model comparisons become issues. Again, having an integrated set of tools makes you faster and less prone to makeing errors. For example, in you're in Pro, you can do factor selection with trees, then go directly into generalized regression and further refine your model. Plus you can use the graph builder and profilers to visualize the solution so that you can communicate your findings effectively.