10 Replies Latest reply on Jan 1, 2014 9:15 PM by Srinivasa Prasanna

# How does one do distribution independent analyses in Tableau?

I am interested to use Tableau to get answers independent of the underlying probability distribution (giving, for example, only the support for bounded distributions).

How does one do this - do we have to write a script?

Thanks

Prasanna

• ###### 1. Re: How does one do distribution independent analyses in Tableau?

An example would be to derive unbiased max/min bounds on data, given samples. For 1-D uniform distributions, this is the classical German-tank estimation.  Does Tableau have support for this and multi-dimensional generalizations, without assuming uniformity?

- Prasanna

• ###### 2. Re: Re: How does one do distribution independent analyses in Tableau?

Hi Prasanna,

See the attached workbook for a solution to the 1D problem. For multi-dimensional problems, so long as max(ID) = count(ID) for the population I'd expect this should also work (note, my data does not have this characteristic).

Hope this helps.

Cheers,

Alan

• ###### 3. Re: How does one do distribution independent analyses in Tableau?

But then we have to explicitly code the extreme value statistical analyses. Does Tableau have inbuilt functions to do this?

- Prasanna

• ###### 4. Re: How does one do distribution independent analyses in Tableau?

Or is there a Tableau package which one can download for this?

• ###### 5. Re: Re: How does one do distribution independent analyses in Tableau?

There is some limited auto-analysis that Tableau will do on data, but I'm not sure it will meet your requirements.

If you are just working with a measure, the Summary card will show an overview of its statistical characteristics:

If you use trend lines to generate a regression analysis, you also get the associated statistics:

Similarly for forecasts:

For more advanced analysis you would either create the necessary functions within Tableau (and it  doesn't have the concept of downloadable packages) or you could leverage the R integration introduced in Tableau 8.1.

Hope this helps.

Cheers,

Alan

• ###### 6. Re: Re: How does one do distribution independent analyses in Tableau?

Thanks a lot - this helped. Instead of R, can we use Java with the Math Package?

Prasanna

• ###### 7. Re: Re: How does one do distribution independent analyses in Tableau?

Not at the moment. The support for R was introduced in the most recent release of Tableau. I'd suggest you either create an entry on our Ideas forum (http://community.tableau.com/community/ideas) or give an up-vote if the idea already exists.

Cheers,

Alan

• ###### 8. Re: Re: How does one do distribution independent analyses in Tableau?

I see that R is not very strong in linear programming and convex optimization. As such, if we need these tools, should we go the text/xml/csv route to Tableau.

• ###### 9. Re: How does one do distribution independent analyses in Tableau?

LP and CP are often used to do these distribution free analyses.

- P

• ###### 10. Re: How does one do distribution independent analyses in Tableau?

My trial ended - how do I extend - I've not used it much.

Thanks

Prasanna