Basic JMP vs SAS overview?

[#1]

Hi,

I was wondering if there are folks who use (or have used) both JMP and SAS, and could give some sort of overview of the difference, and why someone might choose one over the other?  Maybe to keep it simpler, JMP vs PC SAS.

I've been programming in SAS for years, never took a look at JMP.

Recently, I've been working on statistical process control stuff in SAS.  And there's plenty of functionality there (we have the SAS/QC packagakge).  But it seems like every time I see discussion of SPC stuff here on communities.sas.com, it's something from this JMP forum.

And today, saw this great post from John Sall on the JMP blog with a Goal Plot nicely displaying summary information on 100+ processes.

http://blogs.sas.com/content/jmp/2013/09/24/each-statistic-should-have-a-graph-to-go-with-it-not/

And I thought, "Oh, that's beautiful, I should add that to my dashboard..."

And I can do it in SAS, of course, but it made me curious again as to from whence JMP came, where it is going, and JMP vs SAS.

I knew John Sall was co-founder of SAS, but wikipedia told me he developed JMP in the 1980s.

http://en.wikipedia.org/wiki/John_Sall

Which makes we wonder, was JMP SAS Institute's tool for PC analytics before there was PC SAS? Now that PC SAS is around (and EG SAS and web SAS coming....) what does the future hold for JMP and desktop SAS?  Are they seen as complimentary?   Will JMP always be a desktop analytics solution, competing with STATA, Minitab, SPSS, etc?

Would someone looking for an "SPC solution" buy JMP and install it on a few PC's before developing a SAS solution?

Realize this is vague. Happy for any responses, or links to blogs that go beyond basic marketing.

[#2]

I had a few comments in this discussion that might be of interest: https://communities.sas.com/message/181673#181673

[#3]

Here's a view from a relatively new JMP user based on the research I did roughly a year ago when researching statistical packages.

JMP was originally developed for the Macintosh. The idea was to make statistical computing available in the Macintosh style, using a GUI, and to provide a more interactive or exploratory way to analyze data. At that time, the language interface was still the usual way to use SAS -- write some code, run it, look at the result, and repeat. Now, of course, there are several graphical interfaces for SAS, and JMP has a very capable scripting language, so it's no longer just a choice between a language and a GUI.

Apart from matters of style, the biggest differences I found are:

1. If you are just starting out, JMP is probably easier to learn.
2. SAS probably has more total routines.
3. JMP needs all the data in memory, but SAS can process a dataset that's on secondary storage.
4. There is a server version of SAS, so a number of users can share the same environment; and you have greater flexibility to schedule production runs of compiled code using elaborate JCL or scripts -- more like a traditional data processing environment.

I spoke to a number of users of both in my agency, and a few outside. The experienced SAS users felt no need to switch to JMP or add JMP to their kit. The reverse is also true -- the JMP users do not feel the need to add SAS or switch to SAS. There are a few people in both the SAS and JMP groups that also use R, and some more specialized software from EPA and other sources, so it's also fair to say that neither SAS nor JMP serves all needs perfectly.

My decision process was as follows:

1. Both are accepted standards in my agency, and both have experienced users who are willing to help.
2. JMP is somewhat easier to learn and somewhat cheaper.
3. JMP has all the graphical and statistical capabilities I expect to need for the uses I anticipate.
4. I do not expect to need to process more data than will fit in memory.
5. I find the style of the JMP scripting language more congenial than the style of the SAS language.

For these reasons, I chose JMP. The main risk I foresee is that if my organization eventually wants more routine production of fairly standard analyses, we would have to buy more licenses. There's no problem sharing scripts between users, and you can package scripts so the user doesn't have to know the script language, but there are still limits to how automated or scheduled you can make your production environment using JMP. If that gets big or complex enough, you might prefer SAS.

Of course I can't address those of your questions that relate to future product plans, since that's SAS Institute's business, but for the present, JMP is fully supported and under active development.

[#4]

A belated thank you, John.

That is helpful information.  I didn't realize JMP got it's start on Macintosh.  So maybe the key decision for JMP early on was the GUI approach, which SAS came to later (with SAS/Assist, and much later with EG), as well as in-memory processing.

Your mention of people using both SAS and R or JMP and R is interesting, in that it suggests even in an agency which has both SAS and JMP, there aren't many folks who use both SAS and JMP.

Maybe one of these days I'll just download the trial version of JMP and see how it is.  But since I'm now in a place that licenses BI server SAS, not likely that they'd want to start adding JMP licenses at the same time as they are getting rid of PC SAS licenses.

--Q.

[#5]

Quentin,

Like others here my views are similar but here is my readers digest version - JMP is a fantastic tool and very powerful - interactive and quick for data exploration, trying various approaches. I see it as an engineering side of the fence, where you are interactivly developing an analysis and then moving it over to Base SAS or EG for "production".  Obviously, it isn't that cut and dried - JMP does have scripting language and you can do a lot with just JMP. However, as mentioned before, it is memory limited. It can handle much larger files than Excel, but it is still limited to what can be read into memory. But if you have larger datasets, JMP isn't feasible (I have some tables with over a billion rows - not gonna happen in JMP ).

Regards

Fred

[#6]

I have used JMPPro in an academic medical center environment for about 6 years. I support clinical research and clinical trials. Most of the data I see arrives as MYSQL or REDCap databases with the occasional Oracle DB or Excel workbook. JMP and JMPPro allow me to do virtually everything I am asked to do. JMPs' graphical interface makes showing investigators and students how the analysis was done a pleasure - sometimes almost a magical teaching tool. JMP scripting language allows me to provide documentation and code for reporting, but JMP does not easily produce reporting tables of the sort a regulatory agency require. True, with enough scripting, tables can be generated, but it is a lot of work. We re-format the JMP reports by copying them into Excel and using formulas to generate tables. This might work for publication, but not so well for regulatory agencies. Something to consider when you decide between JMP and SAS.

If anyone out there has found a better solution to generating SAS-style tables with JMP, I would appreciate hearing about it.

Regards,

TSP

[#7]

Mr. Parker,

Looking for accessibility information on JMP, I noticed the FDA information at Accessibility | SAS, and it reminded me of your question. Recalling that one can make a data table from a report table in JMP by right-clicking on the report table, and learning from the above that JMP can save a data table in the SAS transport format that the FDA requires, I thought perhaps the two together would save you at least some of the scripting you're having to do.

Best wishes,

John Tate

[#8]

Why not use both? Once upon a time, there was a SAS product supposed to replace JMP, SAS/Insight. It was never really developped and was abandonned a few years ago, leaving JMP as a unique exploration tools until the release of SAS Visual Analytics.

Switching between JMP ans SAS depends on what you are doing, if you are lucky enough to have both licenses:

• With more than a few million rows (around 10 million for a 8GB RAM PC), JMP won't do the job. But, with less, the in memory processing makes it really very much faster.
• Except that JMP doesn't yet seem to use parallel processing on multicore systems, as some SAS procedures do: for some computationaly intensive problems with few data, SAS may run faster.
• If you have to teach statistics, and not programming, then JMP is perfect. I've tried both for years (even decades) and will never come back to SAS for that.
• If your problems are somewhat unique and unstructured, choose JMP to explore your data.
• But if you work in the business intelligence aera, with repetitive problems, complex data preparation, and customized reports with elaborate tables, you wil be much more productive with SAS Enterprise Guide. There is no workflow (or usable track) in JMP, meaning that you can easily forget how you came to a useful result.
• The comparison is the same in the data mining area: JMP Pro is really fast but limited (no text mining, no SVMs), as compared with Enterprise Miner.
• The list of available statistical methods is much shorter in JMP: no mixture of distributions, GEE, bayesian techniques, structural equations, advanced time series analysis, operations research...

Personnaly, I use only JMP with undergraduate students learning statistics, and mostly Enterprise Guide & Enterprise Miner with graduate students learning decision support systems. For research (in business administration), I tend to prefer Guide (with frequent calls to "Open with JMP") for complex but well defined problems, and JMP (with occasional "Submit to SAS") for new problems. For example, these days, I'm working on a Content Analysis problem (coding comics characters): the JMP mosaic plots are really usefull to compare mutiple coders' work, but the Cohen's Kappa in JMP is not corrected for ordinal variables. So I first used calls to the irr R Package from a JMP script or, now and preferably, the Hayes & Krippendorff (2007) SAS/IML macro which computes the Krippendorff's alpha, from another JMP script. The technique was the same for correspondence analysis, until the release of JMP 12.

Hope that helps,

Yves Roy

[#9 · Accepted solution]

Quoting John Sall, Co-Founder and Executive Vice President, SAS, and chief architect of JMP provides a good basic summary:

"JMP is a smaller sibling to SAS, aimed at scientists, engineers, and other researchers who need to analyze data. JMP is to SAS like a spreadsheet is to a database, smaller and geared to interactive desktop uses, but able to merge into the larger enterprise easily. One of the most prominent uses of JMP is to design and analyze experiments. JMP has always been strongest in its graphical approach to analyzing data. There is a graph for almost every statistic, and most of the graphs are interactive.
The largest user group of JMP consists of engineers and statistical support specialists in manufacturing, particularly in pharmaceuticals, semiconductors, chemicals, and consumer products. Often JMP is used in support of a Six Sigma or other quality improvement program. JMP is also heavily used at universities."

We also keep a page on jmp.com highlighting the ways JMP and SAS can be used together.

Daniel Valente

JMP Product Management

[#10]

Hi Daniel,

Thanks for the overview and I had heard similar things comparing SAS to JMP.  I am an old SAS user (I first used it in college around 1981) that has moved to a new job, where they have JMP, not SAS.  I'm used to SAS and writing programs for it.  Is there a good tutorial for getting people like me to make the transition easier and more eficiently?  Thanks.

Tim