His company works with JMP and SAS users to improve their productivity through education and the development of custom applications in JMP Scripting Language (JSL) and SAS. At Predictum’s podcast page, you can set up a subscription in iTunes or get instructions for other pod-catching software. I asked Levin, whose company is a JMP partner, to tell us more about the podcasts.
Me: What is in these podcasts, and who is your intended audience?
Wayne: The podcasts so far have been mostly JMP how-tos, revealing perhaps lesser-known or lesser-understood features of JMP. The audience is composed of both JMP users who are looking to add to their skill sets as well as newcomers and those who are considering introducing JMP into their organizations and want to know more about it.
Please remember that I am a user, not a developer, although over the years, I've had the privilege to interact with just about all of the members of the JMP development team. So, this is my take on JMP.
The one thing that ties all audiences together is the need to understand JMP's design. Analytical applications do not have a common design like, say, word processing applications. This presents both a challenge and an opportunity. If we were talking about a word processing application, or a spreadsheet, there is already a well-established design across a number of products. So, if you are used to using one word processing application, you can probably switch to another without too much difficulty. This may be an advantage, but it is also a disadvantage in that it limits innovation as the product is anchored to one extent or another in the current design.
JMP is not anchored in this way. I've been using JMP since version 3, and even then it was clear that there was a deliberate design that is not like others. When JMP 4.0 came out in 2000, the design distinction became much more apparent. The developers have done a great job not only in alleviating the burden of calculation but also in facilitating the total analytical process.
Me: Can you give us a specific example?
Wayne: Sure. Our very first episode was about Row States, and we explore how it is a "best practice" to keep all of the variant subsets of your data in one data table, instead of having variant subsets of the data across various files (or spreadsheets). Think about it. How easy is it for anyone to pick up a set of data from a project they worked on months ago and run with it? By keeping all the variants in one place (including notes via the column Note property and analyses stored as Table Scripts), it is a lot easier to get reoriented. Keeping variant subsets in one table is easily accomplished with Row State columns. We got quite a bit of e-mail from new users about how useful that feature is and from experienced users who were surprised that after all their time using JMP, they had never explored Row States.
Me: Then the focus of the podcasts is nuts and bolts?
Wayne: So far, but that is changing. Predictum’s motto is "Analytical Productivity." That means that users spend their time getting insights from their analyses, all of them. One of the other terrific things about JMP is how it makes analysis so visual without disqualifying the important role of the raw statistics. I recently read a book by John Medina called Brain Rules. Rule 10 is that vision trumps all the other senses. JMP has some really innovative graphical approaches to analysis. My favorite is the Leverage Plot, and that is the subject of the most recent two episodes.
Me: What has been your most popular podcast?
Wayne: That's difficult to answer. They all get thousands of downloads, but we do get more appreciative feedback on some than others. The Row State episode was certainly popular, but so too was the one on the variety of ways to save reports.
Me: What topics do you have planned for upcoming podcasts?
Wayne: Power and Oneway ANOVA, which will take advantage of one of our set of JSL-based illustrations.
Me: What other goodies do you have at your Web site?
Wayne: We use JSL to develop what we call Analytical Workflows™, applications that are broadly required. We just finished one for Dose Response analysis – now featured on our homepage -- as well as one we call Multi-Response, which facilitates Full and Stepwise Regression across a number of responses (dozens, hundreds, even thousands). It allows users to determine quickly what is statistically significant and practically significant by incorporating specification limits into the analysis.
Me: What is your favorite music to listen to while working in JMP?
Wayne: Almost anything. I love George Harrison's music, always have. From classical rock to baroque to opera -- they all fuel the creative process. Lately, I've been picking up some Tibetan and Indian music.