613251_en-US

liliconcom

Graph Builder Challenge

Hi all,

I feel challenged to make a graph like this:

See here: Graph Builder Challenge 

It is not a particularly nice graph, but can I do any thing similar in graph builder? 

I would also be ok answering with a superior graph 

best

Christian 

 

Re: Graph Builder Challenge

Hi @ChristianBille 

I hate that graph so much. I think it is putting a lot of load on the viewer to resolve those values, and I think separate axes would not only allow for quicker access to the values and comparisons, but also reduce the need to rely on color for identifying the different series. Maybe something like this would work better? You could also use bars, but if someone is making the plot you showed I suspect they're not particularly interested in preserving the link between area and value so a scale starting at 0 probably isn't necessary.  

 

 

 

Re: Graph Builder Challenge

Hi Christian,

Sorry, can't help you with that 2-y-axis-on-the-left chart.

But frankly speaking, it's horrible to understand! Too much superfluous information!

If I follow the recommendations of the 2 books I just finished: 'Effective Data Storytelling' by Brent Dykes and 'Storytelling with data' by Cole Nussbaumer Knaflic, I would propose the following chart.

 

Re: Graph Builder Challenge

I'm very happy that this is a graph that is hard, if not impossible to make in JMP. I like the two alternatives presented here. Agree with @julian , it's a terrible graph.

Re: Graph Builder Challenge

I agree with the others.  It's not a good graph at all!  How about something like this?  It uses the individual data points, but it's pretty informative.

(I couldn't find a good way to get the row state legend into the contour plot.)

 

BMI formula:

 

Profiler if you create a model:

You can see the gradient/slight curvature shown in the profiler as part of the contour plot, as well.

Re: Graph Builder Challenge

It's nice how these alternatives make it so much easy to make the relevant comparisons. Height vs Height, etc. In the original I found myself comparing Height and Weight: "Hmmm, how interesting that height and weight are the same for males."

And I like how Julian's example shows awareness of the variation.

With either of these forms, the axis could include zero so you can graphically compare the relative changes. BMI is not changing as much as the others, relatively.

However, if you rearrange the bar order in the original to put like variables together, you can make the relevant comparisons there, too. It's still messy to figure out which axis to consult. But maybe that's what the original poster had in mind and just shared the first image they found with the basic idea.

Multiple axes in one chart is intentionally built-in in JMP, yet, but here are two techniques that get close. 

1. Multiple charts, side-by-side, in one window:

For this, I made three separate Graph Builder charts and combined them in a JSL New Window/HListBox. You can also get the effect with Combine Windows, but with some dashboard decoration.

2. Normalize all the values to the same scale and graph them in a single chart, but label with a column of values that are in the original scale.

In this case, I stacked the columns and normalized each variable to 0..1 with 1 pegged at 1.2 * mean: 

 

 

 

 

 

Re: Graph Builder Challenge

Hi Christian, 

this is not an answer to your question! Apologize  

 

I agree with you it's not a nice graph. 

On the other way around our customers like these graphs - also with continuous data for process variables. 

During the Discovery Summit several of them asked about the possibility to have graph like the one you posted or with continuous data on the same y axis and different range. 

 

 

Re: Graph Builder Challenge

Thank you all for your response; this was precisely what I hoped for!

His initial post was 

And I felt the urge to stand up for JMPs post on my upcoming webinar! 

I am happy to see that you shared my view on his graph. 

I have borrowed your words and graphs and responeded like this: 

 

and 

@MarcoSalfi 

Multiple y-axis could be used better than the example he has given, and would be cool to just say “yeah we got those!” But then again, it is also good to challenge our customers’ views on what they want versus what is needed for the issue at hand!