11 Replies Latest reply on Jul 25, 2016 7:07 AM by Eleanor Platt

# Parallel coordinates scaling

I am working on a parallel plot visualization and could use some help scaling my factors.

I have attached a workbook with random data that illustrates my issue.

In my case I have several different dimensions, and I would like whatever solution to easily (as it I just drop the dim I care about into the "Color" pill and it works) to change dimensions.  The issue is scaling across the different measures.  I would like to use a % scale (1%-100%) where the 0% is referenced from the lowest aggregate measure for that dimension group, and 100% is referenced from the highest aggregate measure for that dimension group.

For example Meas7 with Dim1 and using a "sum" aggragation has a minumum value at Dim1 value of "Ag" at a value of 1,513,759,750,822 and a maximum at "Ai" with a value of 1,940,867,696,991.  So I would like to scale Meas7 to a 0% is a value of 1,513,759,750,822  and 100% is a value of 1,940,867,696,991.  I also want this to work when I change the color to a different dimension - say Dim6 and when I use different aggragation functions (sum, average, median, count, etc).  For reference I am using the parallel coordinates guidelines from Stephen Few's book "Now you see it" pg 288.

I looked into a table valued function with disappointing results since it operates on the entire dataset, and I wan to operate on the "grouped" data.

Ideas or approaches to address the issues?

• ###### 1. Re: Parallel coordinates scaling

Charles,

You'll need to post a packaged workbook (twbx) for us to take a look.

--Shawn

• ###### 2. Re: Parallel coordinates scaling

As Shawn said, we cannot provide specific assistance without a .twbx, but you may find this example I made some time ago useful:

http://public.tableausoftware.com/views/ParallelCoordinatesExample/ParallelCoordinatesExample

• ###### 3. Re: Parallel coordinates scaling

My apologies.  Here is the packaged version . . .

• ###### 4. Re: Parallel coordinates scaling

Joe,

Thanks for the link.  What I am trying to do is a bit more complex than your example I think.  When I looked through your example I saw that there was no grouping occurring - that each point on the line represents a single value and so the TABLE functions will provide the desired functionality.

In my project I have multiple values that are being aggregated in some way to generate the values that I then want to graph.  It would be like having multiple (different) entries for acceleration for the same car - as if the several cars had been tested and their measures recorded.  So I want to group the low level data into some group based on the dimensions and then parallel plot some aggregation (sum, mean, median, stdev, etc) of that data and since the TABLE functions operate on the entire dataset, and I cannot do aggregations of aggregations I am a bit stumped. . . .

Hopefully that is more clear of what I am trying to do.

• ###### 5. Re: Parallel coordinates scaling

How about something like the attached?

• ###### 6. Re: Parallel coordinates scaling

Joe,

Thanks - it's really close to what I am trying to accomplish.  It took me a while to figure out how to change the dimension being used (when dropped a new dim on the "color" pill I got errors).  So I think to get the full functionality I am looking for I could use a parameter to drive what dimension the Table Value function was operating on.  However, I did not see how to do this.  The only option I saw was to use dimensions.  If you have any idea on how to do this I would be very (even more) grateful.

Charles

• ###### 7. Re: Parallel coordinates scaling

you can use a parameter and a case statement as in the attached to switch out the dimension.

• ###### 8. Re: Parallel coordinates scaling

Joe,

THANKS that is exactly what I wanted.  I appreciate your patience walking me though it and trying to understand what I was trying to do.

Charles

• ###### 9. Re: Parallel coordinates scaling

I suggest that you look at

http://www.amazon.com/Parallel-Coordinates-Multidimensional-Geometry-Applications/dp/0387215077

This book was praised by Stephen Hawking among others. It contains the development of Parallel Coordinates from the

basic definition to the latest results including Topological Approaches for the Visualization of Big Data. Ch. 10 is

full of examples of Visual and Automatic(Classifier)  Exploration of real multivariate data sets (one with hundreds of

variables), query and GUI design and more. There are other applications to Collision Avoidance and Conflict Resolution for Air Traffic Control (3 USA patents), Intelligent Process Control, Decision Support, Multi Objective Optimization.

I'll be happy to respond to questions, remarks, criticism and especially good jokes (see Al-Truisms in my website).

Alfred Inselberg aiisreal@post.tau.ac.il . Note the login is AI IS REAL and AI does NOT stand for Artificial Intelligent

website www.cs.tau.ac.il/~aiisreal/

• ###### 10. Re: Parallel coordinates scaling

Hi everyone,

I went over the thread you have published as far as determining the scaling issue based on different measures but I am not sure I am getting this right with the parameter formula, so I thought posting another request with a simple example attached as twbx.

The question I have is the same, scale the measures from 0 to 100% since I am dealing with ratio from 0 to 1 and quantitative values from 0 to 100.

Thanks very much for your help.

Olivier

• ###### 11. Re: Parallel coordinates scaling

This is sort of the opposite to what you were asking but does anybody know of a way to have separate scales for each variable on Tableau, rather than having to normalise all so they are on the same scale? see here for a picture of what I mean: https://upload.wikimedia.org/wikipedia/en/4/4a/ParCorFisherIris.png

Many thanks.