Is JMP better than R and Python?

[#1]

I've been using JMP since I bought Version One which came on four floppy discs, in about 1995, that was before they had mobile phones.

Never saw a reason to want to consider using anything else, it was heaven then and it's heaven now

I was looking at a Data Science job just now, they told me I had to be able to use Python

I had a look, seems incredibly complicated and limited, I suspect I could do three times the work on JMP as an expert could do on Python at the same time.

Python looks like the old MS-DOS compared to JMP which looks like then Macintosh worked (JMP stands for John's Macintosh Project).

Also, there seems to be stuff you can do on JMP, like neural modeling, that as far as I can see you can't do on Python

Presumably, I can swap databases back and forth between Python and JMP using Excel?

The only advantage I can see about Python is it's free, but I don't care, I'm more than happy with the price I've been paying for a single user license for 25 years.

Did I miss something?

[#2]

The obvious answer is that only you can determine the best tool for your work.

I think that today, more than ever before, there are many excellent tools for data analysis and visualization. I believe that JMP is one of the best. JMP tries to address problems that are encountered by most JMP users when importing data, preparing data, and analyzing or modeling data graphically or statistically. Its current charter, though, cannot address every possible need of all possible users. Even SAS cannot do that! So many versions ago, JMP provided an interface to other software such as Python and R. Users can expand their toolset and take advantage of the collaboration between their tools.

Data scientists are expected to be able to handle any situation, large or small, complex or simple, so having more tools is a good thing. And JMP will continue to grow. You started with version 1. We are now at version 15 with no end in sight.

[#3]

Thank you, but you didn't exactly answer the question if I were a perfectly competent data scientist for 25 years, why do I need to learn Python

[#4 · Accepted solution]

I agree with everything @Mark_Bailey wrote.

Why do you need to learn Python? Having been part of a JMP sales team for a little over 5 years these are the rationalizations people used to NOT consider using JMP.

1. Your organization is frugal with respect to software costs and refuses to pay for JMP. I tried to point out that people's time is a real cost and JMP can often help you discover/decide or take action in a much shorter cycle time compared to coding based software tools. Sometimes people understood this counterpoint...other times...well...not so much.

2. There is some analytics or reporting method that Python is capable of that JMP/JMP Pro/JMP Clinical/JMP Genomics can't do. I agree whole heartedly with Mark that JMP continues to expand capability in an end to end workflow from data acquisition, analytics, reporting, to sharing.

[#5]

Python is a mature, general purpose programming language so you can make your own solution when another software, like JMP, falls short. (You could also use the JMP scripting language to make your own solution in many such cases, too.) Many Python programs and packages have been developed to solve myriad problems that you might encounter, so it can speed up your solution.

I believe that Python was a pillar in data science that was first developed in the computer science realm, not statistics, so it is largely based on the paradigm of 'grow your own.' Data science was the first large scale adoption of machine learning techniques directed towards very large data sets. This effort parallels data mining in statistics but was somewhat separate. They seem to be merging now.

[#6]

I love JMP, but if you try training a massive network with millions, let alone billions of parameters with JSL on a CPU with and no distributed computing of any kind and you will quickly realize why almost any data science job requires Python expertise.

What is your definition of data science?  I'm a statistician by background and I used to balk at the term "data scientist" thinking it was just a sexier word for statistician.  When I really started delving into data science, I quickly realized I was greatly mistaken. 

There is a huge emphasis on AI, particularly through different kinds of neural networks.  Neural networks of today are not the neural networks of the past.  The past decade has seen so many ground breaking innovations, step change after step change enhancement that is very difficult to keep up with all of it.  JMP has a nice basic neural platform, but it would not be practical for most real-world AI problems.  It's more of a classical neural network platform, thought it is capable of having multiple hidden layers.  To be fair, it is not really JMP's niche at all.  SAS Viya has some nice, modern AI tools; but real data scientists need to be able to adapt algorithms to suit the problem at hand.  That's why Viya supports integration with Python.

So to answer your question, you need to learn Python because data scientists are expected to be able to code like nobody's business, understand algorithms at an intimate level and be able to adapt them as needed, and to have tools that can take advantage of GPU and distributed computing.

[#7]

Ditto everyone else. JMP is my daily driver, but there have been some things I've preferred python to JMP that have nothing to do with cost.  

1. Imaginary numbers
2. Tensors (numpy/scipy/pandas in python are awesome libraries)
3. It has more access to OS level functions if you're scripting with a HUGE community of people developing new libraries

My big question is, why not?  It's another tool in the toolbox.  (That being said, I don't particularly like R, so I default to python if I'm not going to use JMP).  

And with the inclusion of JMP's ability to integrate with Python, I can test out what works best for me and still have my nice JMP interface.  

[#8]

I agree with Vince.  It's not a matter of JMP "or" Python (or JMP or R, JMP or Matlab, ...).  It is a matter of "and".  JMP complements many other analytical tools.  When I was using SAS extensively for predictive analytics, JMP was always at my side to help me understand and visualize the data I was using and the results of my analysis.  I currently have a routine type of analysis that I do periodically that utilizes JMP as the interface to R, and it combines analysis from both tools.  

Python is definitely much more of a programming language, but if you have ever used any other programming language, it's easy to pick up and learn.

The reality is that many (most) jobs in "data science" utilize a large toolbox of both commercial and open source tools.    And it is what is commonly taught in many training and university data science curriculum.   I have a goal this year to grow my knowledge and experience in using R and Python, if for nothing else, career security.  I did a search on Linkedin job listing for "python data science" and "jmp data science".  The former had 7000 listings, the latter had 500.  

[#9 · Accepted solution]

Python/R has extreme flexibility in deployment flexibility.  You can make pretty much anything if you have access to the programming resources.

JMP has extreme flexibility in deployed flexibility.  Once you have a deployed something, the user is far more enabled to do their own analytics without having to learn how to code.  This is pretty much impossible in Python/R and results in a ton of heart-ache.  Having to go back to the coder because the manager wants to slightly tweak the analysis is no bueno.

The other thing I'll say is that I believe Python/R to be more fit for purpose for the more exotic machine learning solutions coming out these days.   Principle of Parsimony means these are not actually needed nearly as much as some folks think, but if it is needed you're going to need to look there.

[#10]

YES!

You made a really important point, the people who say you need this or that software skills don't know how to do the job.

Data scientists and AI geeks get data. For doing something creative with that data, i.e. data analytics, in my opinion, JMP is awesome, always has been.