Trellised Bar Chart: data normalization that can help to display more meaningful histograms when one panel dominates the x-axis scale.

Trellised Bar Chart: data normalization that can help to display more meaningful histograms when one panel dominates the x-axis scale.

To enable meaningful histograms, data normalization may be needed when one panel dominates the x-axis scale in a Trellised Bar Chart. 

Histograms can be built by auto-binning the numeric x-axis variable in a Bar Chart visualization, and a multi-panel (Trellised) histogram can be built by using the values of another column in the same data table to define the Bar Chart's Trellis panels.

By design, all panels in a Trellised Bar Chart use the same x-axis scale, and all panels also use the same breakpoints for auto-binning the x-axis data.  

Because of this, if the numeric values in some panels are much larger than the values in other panels, the smaller-value panels can end up with displays that collect all x-axis values into a single bar, as shown in the first image in the attached "Trellised Histogram.docx" file. This can interfere with visual evaluation of the data in those panels.  
  Data normalization can enable display of more meaningful histograms, when one panel dominates the x-axis scale.

As shown in the Cross Table visualization at the top of first image in the attached "Trellised Histogram.docx" file, the largest value in the [Reading] column for a [Measure] column value of "Volume" is many orders of magnitude larger than the largest [Reading] value for any of the other three [Measure] column values (Temperature, TimeSpan, Velocity).

When the Trellised histogram shown at the bottom of the file's first image is built using these two columns, the x-axis scale and auto-binning breakpoints are the same for all four panels, with the result that the first three panels do not display meaningful histograms.

In the file's second and third images, a new calculated column named [NormalizedReading] has been introduced, and it has been auto-binned in the Trellised Bar Chart to define the x-axis in the histogram panels.

This column provides an x-axis variable that has the same range (minimum value of 0.00, maximum value of 1.00) in all of the Trellis panels.
 
The following expression defines the [NormalizedReading] calculated column for this example:
 
([Reading] - Min([Reading]) OVER ([Measure])) / (Max([Reading]) OVER ([Measure]) - Min([Reading]) OVER ([Measure]))
 
Tooltips have also been defined, to allow the user to examine the original [Reading] values that correspond to the edges of the [NormalizedReading] bars in each panel's histogram.  The file's third image displays the Tooltip for one bar in the Velocity histogram.

~~~

Note:  Existing enhancement requests have asked for implementation of the following user options in a future Spotfire release:

  - independent x-axis scales for each of the panels in a Trellised Bar Chart
  - independent x-axis auto-binning breakpoints for each of the panels in a Trellised Bar Chart


You can also vote for this existing suggestion on the TIBCO Ideas Portal:

https://ideas.tibco.com/ideas/SPF-I-2740
(Multiple x-axis scales in trellis)

You can also suggest new ideas (such as other enhancements to the Trellis feature) on the Ideas Portal, at this URL:

https://ideas.tibco.com
?