I was under the impression that R was more flexible than Tableau, but a steeper learning curve, or that any chart Tableau can make, R can make as well (maybe sans interactivity), but with the added requirement of understanding a programing language. I have only seen impressive things people have made with R, and not actually developed anything in R yet, so my opinions of R may be flawed.
What would be a use case where someone would build an analysis in R and then want to run the resulting data through Tableau?
If you are able to automate/schedule the analysis of data in R, can you not also schedule the extraction of that data out of R and into a data source that Tableau can connect to, or use some kind of ETL tool?
I have thought of R to be at a similar level in the cycle of analysis as Tableau, that while solving different problems and both useful, that they sit parallel to each other, that your data is prepared for analysis in either Tableau or R (or both), but not Tableau and then R or R and then Tableau. I do not view either application as a data transformation application for processing and loading into another application, but rather I consider both to be final data presentation applications.
In other words, I personally would not consider either to be middleware. I prefer to use dedicated ETL applications for transforming data, enabling greater analysis capabilities in applications like Tableau and R.
What is your viewpoint on these applications? do you use both?
I use R and Tableau together and would also like to see Tableau be able to work with R. R can do more types of data manipulation and statistical analysis than Tableau, but Tableau provides a much more fluid means of interacting with and examining data.
The way I use R and Tableau together often goes something like this:
1. Pull a dataset from a database such as Oracle and explore it using Tableau. Often that leads to refined queries to deal with data quality problems that Tableau is very good for uncovering.
2. Read output from the refined query into R, and crank some statistics. The output is often tabular, based on R data.frames. For example, if you were looking at baseball players, you might use R to generate a set of performance statistics and perhaps predictions for each player's end-of-season batting average.
3. Write R data out to a .csv or a relational data table, and read it back in to Tableau to explore the results and produce a set of reports.
If Tableau knew how to read in R data.frames, that would save steps. And if R knew how to deal with Tableau extracts, that would both speed up some analyses and help R scale to larger datasets (R works principally in-memory)
My thoughts and interest in this are very much like 'guest'. The benefits of using R from within Tableau would be to provide the ability to use R's statistical functions within an interactive visualization.
Let's say a user was exploring a dashboard and selects a subset of the total data that was of interest. By using R from within Tableau then R could take that subset of data as a table, do some calculations and return the results to Tableau for integration into another part of the dashboard. It would be very similar to using calculations in Tableau, but using R commands instead of native Tableau calculations.
My understanding is that you can do similar things using SAS, SPSS and even Excel (using RExcel add in)
Here is a bit of info and an example of how it is done in Spotfire:
Obviously Tibco own S but a similar method using R would be superb.
I cast my vote for this feature. It would make both programs better. R is THE statistical program. Notice how many programs in the recent past have developed some form of integration with it.... Even SPSS :-)
I would just like to add my support for this idea. I use R extensively and I'm just about to embark on my first Tableau project, the reason for this is they have very different strengths. Tableau's great advantage seems to be interactive graphical data exploration. I have used several of R's interactive data visulisation packages however none, at this time, seem to offer the same quality of graphics as Tableau. This is in strong contrast to static graphics in which R can more than compete I believe.
However, I will certainly be using them together. Firstly I will pre-process a great deal of data in R (in a dashboard comparing various stocks I might compute the co-integration quality with the Index or a structural change indicator and so on...). At the moment I will achieve the connection to Tableau by outputing in the appropriate format and reading into Tableau, I do not mind this too much. The second aspect provides more of an opportunity to my mind, the ability to add a model fit to a scatter plot from example where the model is calculated in R would be invaluable - an would greatly increase the amount I would use Tableau I think.
What is more such integration has been managed with a number of other systems (see above but also Mondrain + GGobi + GTK based systems), I realise that Tableau probably has its reasons for remaining seperate but I think for a large community (as Ernesto mentioned R is THE statistical program) such features would be an incredible addition. As Andrew mentions even excel can basically do the above (via RExcel & RCom) and I'm no fan of excel!
One final example: I would love to be able to dynamically select data in the tableau interface (say on a scatterplot - or a bar chart linked to a scatterplot) and for R to recompute a model for this just subset and update the fit line etc.on the plot in real time. I've seen it done in other systems (java based iplot if i remember correctly) and it is SO useful!
I would also feel very pleased to see that connector available soon!
Melding R and Tableau would be a dream come true for me. I don't really see this happening in the same fashion as other tableau data "connections." R is not a database, and as others have pointed out, you can already interface between the two through the cumbersome step of exporting data from either program.
What we really need is the R console functionality inside of Tableau. Highlight data, right click, and "copy data (or crosstab) as data frame". Or some kind of "use R to add rows" functionality. Here's what I do now: 1) pull data into Tableau, 2) clean, filter, and structure data, 3) Copy data to excel, 4) pull data into R, 5) run a variety of time-series forecasts, 6) export data to .csv, 7) clean/combine/categorize results in excel, 8) pull data back into tableau for presentation.
The R forecast package has a simple forecast() function. How about this: highlight data to use in model, right click "forecast", "How many data points?", "10", presto-more data (with a new category created to distinguish it from the actuals). That alone would be a benefit. It's all open source... why not leverage it?
Tableau computes trend lines, but really, what's the point if you can't use them to create forecast data? It's just a pretty line.
This is a great idea. Both products have great qualities, but R stands out in the statistics and modeling realm...Tableau interactive dashboarding capabilities along with Server for a BI platform make it better for making the information available and consumable.
In Tableau 8.1 you can create calculated fields using R. If you're participating in Beta, you don't have to wait for the RTM, look for new functions with names starting with SCRIPT_
You can integrate Tableau and R in following steps:
Step1：Install Tableau 8.1 and R ，you can get install files from ：http://cran.r-project.org/
Step2： Install R server ，start R and type:
Step3：Tableau connect R server
Open Tableau desktop ,click Help ——Setting and Performance ——Manage R connection
The correct R command is
, which would result in error message
package ‘Rserver’ is not available (for R version 3.1.2)