One could write 40,000 words on this subject. Is there a specific statistical or computer issue that is of major concern to you? It would be easier to answer questions about specifics than comparing these software packages in their entirety.
I guess I'm after breadth rather than depth. That is:
1) Are JMP and SAS complements to one another or is JMP a niche product and SAS is the main stats program?
2) I found a thing online comparing very old versions of SAS & JMP (ver 6 and 3 respective, if I remember correctly). Their take was SAS is "programmatic" where JMP was "point-and-click".
3) One thing I found today going through some promo material, which was interesting to me since I'm also learning C#/.NET, is that SAS has a .NET framework through which you can access most, if not all, of SAS. I infer that JMP does not. Also, if you need custom or exotic models or have extremely large data sets (>10^6) then SAS is your only option.
Is #2 correct? What are your thoughts on question 1?
1) JMP and SAS can be complimentary to one another, or they can stand alone. There is a fair bit of overlap in the capabilities of JMP and SAS. I would not call JMP a niche product; JMP is a powerful statistical package.
2) JMP now has a programming language, so it can be used in "point-and-click" mode, or via programming commands; however I dislike the programming language in JMP and avoid using it as much as possible. SAS can be used in programming mode, or there are full-screen GUI interfaces that can be used.
3) SAS has more modeling algorithms available. JMP can handle some very large data sets, I have not yet reached a limit. Regarding "custom" models, JMP can fit some "custom" models, don't really know what you mean. I don't use .NET so I can't answer those questions.
I do nonlinear regression and work with folks using SAS 9.2. I strongly prefer graphics in JMP with superior GUI controls (much faster to make the graph pub ready). With good and sloppy data, our regression results agree as much as nonlin results can. JMP can do some more complex stats too, e.g., JMP can also do iteratively reweighted regression, which I had previously thought would have required SAS NLIN. JMP is highly interactive, so you can explore (and play around with) data and platform output much more easily and faster in JMP than in SAS. Also- check out the JMP website for how JMP can be the front end for some SAS procs. My recommendation: use both.
A bit of history perhaps....SAS is the parent. At one point a gentleman (whose last name began with the letter J) wrote a program for Macintosh Platform...hence the acronym JMP. It expanded into the pc market after being successful for mac. SAS has many more capabilities...enough said.