Sounds like a great idea...anyone know of someone who's knowledgeable of both products?
I actually have some familiarity with both R and Tableau. I spent the last 3 months shopping, demoing, and doing sales calls on all my options before I really dedicate myself to complete mastery or anything. I ended up concluding what a couple crack Berkeley CS friends of mine thought an analyst should do (Tableau and R).
If you are familiar with any basic "programming" in other command line style 4th generations languages (high level) like SQL, SAS , STATA, etc., then learning R would not be that difficult. All I had was a couple intro CS courses in undergraduate, and I had no trouble learning STATA for model building in graduate school for example. R is definitely a natural compliment to Tableau and helps fill in many of its current serious statistics, modeling, data mining limitations. Tableau really needs to build R integration or do all the features for themselves. If they don't do that by version 9 at the latest, I will be very disappointed. TIBCO Spotfire has had it for years.
I am actually trying to put together an integrated "analytics stack" as I like to call it consisting of:
Tableau Desktop Professional ->Reveloution R Enterprise->Mathematica 9
Mathematica 9 has R integration. Now I just waiting for Tableau, so I can seemlessly move between the 3 depending on what I want to accomplish. I really can think of few things under the sun I could not accomplish between the 3 for my purposes. And they would all still cost less than the first year of SAS.
I am actually becoming a big fan of R and one commercially supported version in particular, Revolution R Enterprise, which can handle large datasets as effortlessly as SAS at a fraction of the price.
Still, I would just download their free version basic version and the Rcmdr and Rattle GUI packages to start learning if this is something you would find valuable. There are also a number of good books for starting out with R I have found. R Cookbook, R in a Nutshell, The Art of R Programming, Data Mining With Rattle and R-The Art of Excavating Data for Knowledge Discovery, etc.
Hope you all found this helpful!
It sounds like you have great experience integrating other applications with R.
Perhaps, you would like to share your ideas for relatively seamless yet flexible interface between R and Tableau on the idea page for this topic.
We are starting to plan the next release and your ideas would be super helpful.
John - I'd love to see a WebEx with some examples of how you are using Tableau with R today. Would you be interested in sharing at one of the TDT sessions?
@ Scott. Thanks! I am very heartened to hear this is an internal priority at Tableau for the next release (ETA?). You look to be the exact guy I have been looking to get in touch with at Tableau on this. I would be glad to think about my current typical workflow for moving between programs and brainstorm a bit about how the integration might look. Glad to correspond with you at: email@example.com
@ Mark. Unfortunately, I am a bit pressed for time today. Also, I don't think that would make for too much of an engaging demo of me manually exporting and importing between the two programs, lol. One day though...One day!
No promises or ETAs. But, the execs have certainly noticed the impressive number of votes for R Integration.
I'll be swamped finishing up 8.0 this month.
After that, I'll spend time thinking about possible future statistical features including R integration.
To integrate with Tableau zen, the details are important.
Just to clarify, Spotfire & R integration is not as simple as hooking up a desktop product to open-source R. Running R scripts in Spotfire requires TIBCO Spotfire Server (TSS) and TIBCO Spotfire Statistics Services (TSSS) licenses as well as the TIBCO Spotfire Web Player. I have not used it for R but with SAS in their DecisionSite product. Even that required some programming experience in both Spotfire and SAS.
If Tableau were to provide a mechanism to hook up with R from the desktop.... well, I think that would be Tableau Zen indeed.
@ Scott. Thanks for filling us in on the inside skinny. I actually just upgraded to Mathematica 9 which now includes R integration of their own, and they describe their method as "using JLink and RJava / JRI Java libraries to link to the R functionality" but they do have some performance cautions with their method "RLink is a rather high-level interface, built on top of JLink, which itself is built on top of MathLink, and RJava/JRI, which is a Java interface to R runtime."
Thanks again for all your hard work and we will definitely look forward to the forthcoming quant/stat zen treatment.
@ Tim. Yes, thanks for that clarification. You do need those add on licenses, and it is definitely not a seamless process by any stretch. I thought it was a bit more straightforward with S+, their commercially supported R equivalent.
John - I wouldn't think you'd demo the import / export process. I'm personally interested in seeing live use cases with a high level overview of how the process works and the best place to find resources to learn how to do it. Think like a cooking show. They show you the ingredients and the steps, then pull the finished product out of the oven.
This sounds like a great idea!
At the moment the vast amount of my R / Tableau activity is in the area of preparing new tables in R before bringing it into Tableau. Part of this is simple data transformation work, eg using the reshape package but I'm increasingly using it for more sophisticated uses, eg generating x / y coordinates to plot network diagrams. Of course the limitation is where calculations need to be made after a users' interaction. Sometimes you can predict all possible selections and provide 'answers' to these.
I wrote a high-level case on adding prediction into dashboards here: http://www.organizationview.com/predicting-hiring-success
I'm increasingly interested in visualising 'difference from expected' which of course requires a calculation of what is the expected value.
The final area where I'm increasingly using the two together is by prototyping in Tableau which I'll then build in R. At the moment this works well when the final output is printed (e.g. personalised reports / presentations). The pdf output I get from R is much cleaner than Tableau (even seemingly v8) meaning I can do more getting it perfect in something like Illustrator.
Andrew - that's a cool article, thanks for sharing. I'm exploring some of the other pages now and they look great.