topic Re: Graph builder: issue with multiple response in Discussions

Graph builder: issue with multiple response

hogi — Wed, 21 Jun 2023 08:30:02 GMT

Since Jmp 17, Graph Builder can interprete columns with modeling type Multiple Response in a way that the entries are split and the values show up one by one. Amazingly useful!!

But there seems to be an issue if the column is used in different drop zones.
If the user wants to find out which of the students play multiple sports , the result looks boring:
there is just a diagonal line :(

Seems that Jmp splits the students into smaller entities with a well defined sports activity.
So Katie, who plays Soccer and Tennis becomes Katie1 (soccer) + Katie2(tennis) without a link between Katie1 and Katie2.

Is this issue already known and will it be fixed in Jmp 17.2?

Is there a workaround in Jmp17.1?

dt = Open( "$SAMPLE_DATA/Big Class Families.jmp" );

dt << Graph Builder(
	Variables(X( :sports ), Y( :sports )),
	Elements( Heatmap( X,Y ) )
);

Re: Graph builder: issue with multiple response

Ross_Metusalem — Fri, 23 Jun 2023 16:35:46 GMT

Agreed on both fronts: Improved multiple-response handling is a welcome development, and this certainly isn't a particularly informative heatmap. A couple thoughts:

I wouldn't say that Katie is now Katie1 and Katie2 without any link between them. If you select Katie in the data table, both of the corresponding cells in the heatmap are highlighted. So Katie's data have not been split in the strict sense of creating multiple observations out of one. But you're right that a single row of the data table is being represented across two different heatmap cells.

I imagine you're looking for a heatmap that would color a single cell for Katie at the intersection of Soccer and Tennis (but please ignore the rest of this if I've misunderstood). Note that creating such a heatmap isn't as trivial as it might seem, because the multiple response column allows for more than two sports. For example, Jaclyn plays Basketball, Water Skiing, and Volleyball. There isn't a single cell that could represent the intersection of all three sports, so Jaclyn would still need to be "split" across cells in the heatmap in some way.

Re: Graph builder: issue with multiple response

hogi — Fri, 23 Jun 2023 16:52:14 GMT

Hi Ross,
thank you for your support. I will write a wish list and ask if it's possible to adjust the behavior ...

@Ross_Metusalem wrote:
I imagine you're looking for a heatmap that would color a single cell for Katie at the intersection of Soccer and Tennis

yes.

@Ross_Metusalem wrote:
For example, Jaclyn plays Basketball, Water Skiing, and Volleyball. There isn't a single cell that could represent the intersection of all three sports, so Jaclyn would still need to be "split" across cells in the heatmap in some way.

I would expect that Jaclyn counts "1" at the intersections of:
- Basketball - Water Skiing

- Water Skiing - Volleyball

- Volleyball - Basketball

... and the same flipped in the other half of the matrix. So, no big issue.

If there are 10 students playing soccer and tennis, they will count 10x "1" at the intersection of:
- tennis and soccer.

But you are right. If sports activities show up at some more locations in the graph (Group X, Page, etc. ), this will soon or later crash the system if all possible "intersections" are tracked.

For such a case, Jmp could give a warning - e.g. not more than 3 drop zones filled with multiple response values.

Re: Graph builder: issue with multiple response

Ross_Metusalem — Fri, 23 Jun 2023 17:07:05 GMT

You're right, it's not hard to imagine the heatmap behaving that way. It could get tricky to interpret if there is strong imbalance in the number of responses across observations. For example, if most students play 1-2 sports but a few play 5+, those students sports pairings will dominate the heatmap (and perhaps swamp it out entirely) and perhaps create a false impression about the diversity of pairings across all individuals. Just something to consider.

Re: Graph builder: issue with multiple response

txnelson — Fri, 23 Jun 2023 17:21:47 GMT

I believe the chart you want is Person by Sport

In the chart below, you can see that Jaclyn is in 3 sports.

Re: Graph builder: issue with multiple response

hogi — Mon, 02 Sep 2024 20:29:35 GMT

Actually, a graph like this one:

The diagonal line shows the number of students participating in each sport, and the off-diagonal entries shows which sports are listed together and how often.

Re: Graph builder: issue with multiple response

hogi — Mon, 02 Sep 2024 20:36:44 GMT

maybe more interesting: visited countries.
cells with black outline: Kirk - he even visited Antarctica!!