Bivariate - Porting Fit Equation to a Data Table

[MShroff]

Hello,

The Fit Y by X Bivariate (and by extension, Fit Model) platform currently does not have the ability to send the fit equation (aka, prediction expression in Fit Model) to a data table. IOW, when right-clicking on the equation, the menu does not have commands for Make Data Table or Make Combined Data Table, which are available for the other blocks.

It would be helpful to add this feature to be able to summarize the equations into a table. When several similar reports are generated (e.g., with the use of By variables or when there are multiple X or Y variables), it's not feasible to copy/paste or manually type them into a spreadsheet.

Regards

Mehul Shroff

[Sarah-Sylvestre]

Hi @MShroff, thank you for your suggestion! We have captured your request and will take it under consideration.

[SamGardner]

@MShroff could you share more information about what type of problem you are working on when you want to do this?  Also, do you have any suggestions for how the output results should look?  

You can save parameter estimates for fitted models to a combined table for both Bivariate and any of the model fitting platforms, and with some relatively straightforward table manipulations get to a table of parameter estimates organized in whatever way you want.  Does that meet your need?  

[SamGardner]

Also @MShroff , there is a JSL function, Summarize YbyX( ), that might be helpful for what you want to do

Names Default To Here( 1 );
Open( "$SAMPLE_DATA/Lipid Data.jmp" );
Summarize YByX( X( :age, :Weight ), Y( :HDL, :LDL ) );

which create a summary table of the regression coefficient estimates and fit statistics as shown below

[SamGardner]

Also you can use By variables by specifying them with a Group( ) argument to the function

Names Default To Here( 1 );
Open( "$SAMPLE_DATA/Lipid Data.jmp" );
Summarize YByX( X( :age, :Weight ), Y( :HDL, :LDL ), group(:gender) );

[SamGardner]

Also, the JSL is just a short-cut to get the final result table that you can generate from the Response Screening Platform (Analyze > Screening > Response Screening) 

[MShroff]

Hello, @SamGardner,

I'm simply trying to write out the equation to a column in the data table. I understand that the parameters can be saved but for documentation / communication purposes, having the ability to tabulate the full equation would be helpful.

Regards

Mehul Shroff 

[Jeff_Perkinson]

Thanks, @MShroff.

It would be helpful if you could post an example of the data table you'd like to get from this ability. I'm having a hard time visualizing exactly what you want. 

[SamGardner]

Thanks again for the input and idea.  We have decided not to pursue this idea, but please let us know if you have any other input or ideas on how to make JMP better.  

[JumboFiatco]

For simple Bivariate fit (no BYclause):

// Get Fit Equation
		report = Report( bvar ); // bvar - reference to the plot
		eq1 = report[Outline Box( 2 )][1, 1] << Get Text; // "First item of the first item of the second outline box"
		eq2 = report[Text Edit Box( 4 )] << Get Text; // Direct address. Numbering not always predictable.
		Show( eq1, eq2 );
// Annotate Fit Equation back to the plot
		report << Add Text Annotation(
			Text( report[Text Edit Box( 4 )] << Get Text ),
			Fixed Size( 0 ),
			Text Box( {xdim_bivar * .25, ydim_bivar * 1, xdim_bivar * .25, ydim_bivar * 1} ), // location arguments
			Font( "Arial", 10, "Plain" )
		);