Parallel Plot in Graph Builder - how to get the same scale for all values

[#1]

Parallel Plot is very useful to analyze repeated measurements and show the bin flips.

Unfortunately GraphBuilder scales the axis for each value separately. Therefore, one and the same bin is at different heights.
This complicates the analysis of the plot.

Is there an option in Jmp to synchronize/link the axes between the different parallel values, like it's possible to link axes between different pages?

[#2]

I can't reproduce a graph like yours without a version of your data but try right clicking on the axis label and trying some of the other Combine Scales options to see if one of them gives you what you're looking for.

[#3]

Hi @Jeff_Perkinson , thanks for helping to find a solution
Combine Scale sounds like the right place to look at ...
Unfortunately, none of the generated plots looks like expected, i.e. same values on the same height, indicating 
horizontal line ~ no bin flip

Here is a code to generate a test data table:

View more...

dt = new  table();
dt << add rows(10);

dt << new column("meas1", ordinal, set each value(randominteger(4)-1));
dt << new column("meas2", ordinal, set each value(:meas1));
dt << new column("meas3", ordinal, set each value(mod(:meas2 + if(randominteger(10)<5,randominteger(2)+2,0),4)));
dt << new column("meas4", ordinal, set each value(mod(:meas3 + if(randominteger(10)<5,randominteger(2)+2,0),4)));
dt << new column("meas5", ordinal, set each value(mod(:meas4 + if(randominteger(10)<5,randominteger(2)+2,0),4)));

Graph Builder(
	Variables(
		X( :meas1 ),
		X( :meas2, Position( 1 )),
		X( :meas3, Position( 1 ) ),
		X( :meas4, Position( 1 ) ),
		X( :meas5, Position( 1 ) ),
		Color( :meas5 )
	),
	Elements(
		Parallel( X( 1 ), X( 2 ), X( 3 ), X( 4 ), X( 5 ), Legend( 3 ) )
	)
)

[#4]

Hi @hogi,

With nominal/ordinal variables, I don't believe there is going to be a way to force the Y location to be identical for the same value levels because with those modeling types Graph Builder will attempt to show proportionality of each level at each variable, which for any real-world case means variability in that Y location across a set of variables.

If you want to ensure the same axis for all variables in a parallel coordinate plot, you'll want to have your variables as continuous (and a proxy variable if you're using one of them in another role, like color):

If the scales are very different, you will also need to instruct Graph Builder to always merge parallel axes, which is an option you can find under the Red Triangle while holding shift:

If you select "Always" for Parallel Axis Merging, then regardless of how different the scales are, Graph Builder will use a single scale for the Y axis rather than implicitly standardizing the space.

I hope this helps!

@julian 

[#5]

Thanks @julian for the explanation.

The disadvantage of continuous variables vs. nominal/ordinal ones:
the lines are put on top of each other, so the graph will look the same for 1 and 1000 units with bin flips.

Here a comparson of both approaches - I really like the one with nominal values:

I hope other Jmp users have similar "niche application" and will support this wish:
Parallel Plot: synchonized axes 
to add an option to synchronize the scales

[#6]

@julian wrote:

instruct Graph Builder to always merge parallel axes, which is an option you can find under the Red Triangle while holding shift

---

another easter egg for the list
CTRL/Alt/Shift + click/select/double click/right click 

[#7]

@Lisa_Daroque 

Parallel Plot: synchonized axes  could be very helpful for Semiconductor applications.

The idea:

drag multiple columns with 0/1 (...)  values onto x axis and activate parallel plot

--> check for bit flips between the columns:

most important:
the width of the lines:  more devices -> thicker line.

a few devices -> thin line.

The huge disadvantage at the moment:

the "random" position of 0 & 1 on the y axis !!!

@julian 's suggestion fixes THIS issue, but makes the graph useless - the cool "width" effect is gone