nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Better way to plot multiple FitY by X plots

OneNorthJMP

My original source table is the tall table with ID, group and tons of params. I want to plot all params Fit Y by X (Group A vs Group B group by ID). The traditional way i do is i change the tall table to wide table and plot one params at a time. But if i have hundreds of params, it is not a good solution. Is there any better or smarter way to do this ?

attached sample jmp files

tall table

switch to wide table

Plot paramA A vs ParamA B

DEBUG.jmp

Accepted answers

txnelson

Taking your example data, I first split the data into your Wide form, and then I took that data table and did a multi series stack

Tables=>Stack

specifying a 2 column stack

I then added a simple new column called Analysis which is a simple concatenation of the Label and Label 2 columns

Finally, I ran the Fit Y by X specifying Data and Data 2 as the Y and X columns and then setting the Analysis column as the By column, giving the results below

All comments

txnelson

Taking your example data, I first split the data into your Wide form, and then I took that data table and did a multi series stack

Tables=>Stack

specifying a 2 column stack

I then added a simple new column called Analysis which is a simple concatenation of the Label and Label 2 columns

Finally, I ran the Fit Y by X specifying Data and Data 2 as the Y and X columns and then setting the Analysis column as the By column, giving the results below

OneNorthJMP

Hi Nelson,

How you do the 1st step - specifying a 2 column stack? Can you further elaborate?

ian_jmp

@txnelson is referencing the 'multiple series stack' option:

Screenshot 2020-11-25 at 15.06.32.png

KarenC

Another platform that you might explore is response screening. I find it a helpful tool for looking at many fit y by x's. See response screening documentation .

Same concept as @txnelson but I would arrange the table just a little differently by stacking to make a very long table, then splitting by the group column. Finally use the parameter in the by column of a bivariate script:

names default to here(1);

Long = Data Table( "DEBUG" ) << Stack(
	columns( :paramA, :paramB, :paramC ),
	Source Label Column( "Parameter" ),
	Stacked Data Column( "Data" )
);

repivot = Long << Split(
	Split By( :GROUP ),
	Split( :Data ),
	Group( :Parameter, :ID ),
	Sort by Column Property
);

biv = Bivariate( Y( :A ), X( :B ), Summary Statistics( 1 ), By( :Parameter ) );

dtSum = (biv << XPath("//TableBox"))[1][1] << make combined data table;

dtSum << Select Where(:Column 1 != "Correlation");
dtSum << Delete Rows;
dtSum << Set Name( "Results" );

Then the resulting table can be sorted to find the interesting correlations.

The response screening platform certainly seems promising though.