Share your ideas for the JMP Scripting Unsession at Discovery Summit by September 17th. We hope to see you there!
Choose Language Hide Translation Bar
Highlighted
spinup
Level II

How to Plot graph with reference lines for each element of table summary

Hi All,

I am trying to  plot graphs  (script , JMP 13) with reference lines for each element of my table summary (graph of score vs # try for each Name+Grade, then add reference line of 25 and 75 percentile of Score on x axis). Would appreciate any comment 

NameGradeScore# try
name1A80

2

name1A823
name1B983
name1C862
name2A883
name3A902
name3A881
name3A943
name4B962
name4B824
name4A1001
name4C922
name5A843
2 ACCEPTED SOLUTIONS

Accepted Solutions
Highlighted
spinup
Level II

Re: How to Plot graph with reference lines for each element of table summary

Thanks @Craige_Hales for the link, but my main difficulty is how to program to output the individual graphs as each have their own reference lines, more clearly, I am wondering how to generate graphs for each color in this data table :

jmp question-2.jpg

Thanks!

View solution in original post

Highlighted
DS
DS
Level VI

Re: How to Plot graph with reference lines for each element of table summary

Hi @spinup,

 

  I think I understand what you want. But, I'm also a little confused: are you wanting to summarize or just graph the 9 separate bivariate graphs?

 

  Working off your original data table, you can easily create all 9 graphs if you do a bivariate fit, e.g.:

Bivariate( 
  Y( :Score ), X( :Name( "# try" ) ), 
  By( :Name, :Grade )
);

  This will generate your nine graphs that you had colored in your previous post.

  Then, what you'll probably have to do is use a For Loop(s) to generate the quantile lines (or shaded area).

  The script for the first graph would be something like:

Bivariate(
	Y( :Score ),
	X( :Name( "# try" ) ),
	Where( :Name == "name1" & :Grade == "A" ), // this would be selecting 
	SendToReport(
		Dispatch(
			{},
			"2",
			ScaleBox,
			{Min( 75 ), Max( 100 ), Inc( 5 ), Minor Ticks( 1 ), // this sets the appearance of the y-axis
			Add Ref Line( 83, "Solid", "Blue", "25th quantile", 1 ),
			Add Ref Line( 95, "Solid", "Blue", "75th quantile", 1 )}
		)
	)
);

  In fact, if you set up a list of the names NameList={"Name1", "Name2"...}, or pull it from the column elements, then you can have a very short script by putting in For Loops and just us reference the NameList[i] element. This would be also the case for the Where(:Name==NameList[i] & Grade == GradeList[i]).

 

  Another option would be to inspect the tree structure of the output from the first script I included above and then you'd reference the AxisBox(i) for each of the y-axes for the 9 graphs. The y-axes are all odd, so if you do a For Loop, then you'd have to do i++ so that it iterates from 1 to 3 to 5 and so on in the loop.

 

  There's some really good stuff in the link that @Craige_Hales provided.

 

  For your specific issue, I think you'll definitely need to do some looping. Either that, or you can do the whole thing manually once and then just save the script to the data table (probably as a group), and then you can copy/paste that and save as a separate JSL file in a new script window. If you can figure out how to do it manually, JMP will write the JSL for you, and all you need to do is put it all together into a master JSL file that takes care of it all for you. The only issue there is if you decide to change names or column references or things like that. If you do, then you'll need to start using local variables.

 

Hope this helps!,

DS

View solution in original post

6 REPLIES 6
Highlighted
DS
DS
Level VI

Re: How to Plot graph with reference lines for each element of table summary

Hi @spinup,

 

  I'm not sure I understand entirely what you want to do, but you might have a hard time scripting the whole thing entirely. If you are keeping the data table around for a while, or load it in, you might want to do the analyses separately and then save them to the data table.

 

  If you do want to script the whole thing, you'll need to still run the analyses to determine the 25/75 quantiles, etc.

 

  When you say you're graphing "score vs # try", do you mean when you do a "fit y by x" that score is "y" and "# try" is "x"?

 

  Also, when you say table summary, do you mean you're finding the mean of score and #try while grouping by name and grade? Or, do you mean you're tabulating by name the grade and finding the mean of score & #try values?

 

  You could do it a couple ways though. I took your table and brought it into JMP, and here's what I did (assuming I am thinking about it the same way as you):

Tabulate by Name (A, B, C are the # tries), this helps to get the means, but you can script this separately if needed.

Snap1.png

With this at hand, I graphed it a couple different ways: bivariate fit or histogram (histogram already has the quartile info readily available from the box plot). and then you can add reference lines via the normal way.

Snap2.pngSnap3.png

 

For scripting all this out, be sure to use the red hot button to get to the actual JSL script to generate what you want. But again, if you save the script to the data table, then each time you open up the data table, you can click the green hot button and re-run it quickly.

 

Hope this helps!,

DS

Highlighted
spinup
Level II

Re: How to Plot graph with reference lines for each element of table summary

Thanks @DS  for your reply and sorry for not being very clear. Here I would like to explain a little more:
first of all, I am trying to write the script in a separate script window so I can run it independently, here is the first part of the script where I am trying to get the summary:

Names Default To Here( 1 );
dt = Current Data Table();
Summary = dt << Summary(
	Group( :Name, :Grade ),
	Median( :Score ), 
// other statistics can be added here//
	Link to Original Data Table( 0 )
);


the Output will look like this :

jmp question.jpg

Now , What I really want is the 9 graphs ( Score vs # try) (one for each row (having N Rows datapoints) ), and then use the statistics of each group to add to the reference lines ( all by scripting). 

In your graph example, looks it is for name1 with any Grade.

Thanks

 

Highlighted
Craige_Hales
Staff (Retired)

Re: How to Plot graph with reference lines for each element of table summary

Highlighted
spinup
Level II

Re: How to Plot graph with reference lines for each element of table summary

Thanks @Craige_Hales for the link, but my main difficulty is how to program to output the individual graphs as each have their own reference lines, more clearly, I am wondering how to generate graphs for each color in this data table :

jmp question-2.jpg

Thanks!

View solution in original post

Highlighted
Craige_Hales
Staff (Retired)

Re: How to Plot graph with reference lines for each element of table summary

Maybe something like this

dtBigClass = Open( "$sample_data/big class.jmp" );

dtage = dtBigClass << Summary( Group( :age ) ); // makes a small table of unique ages
ages = dtage[0, 1]; // [12, 13, 14, 15, 16, 17]
Close( dtage, "nosave" ); // table no longer needed

grid = Lineup Box( N Col( 3 ), spacing( 20, 30 ) ); // Horz,Vert extra space between graphs in pixels
For( iAge = 1, iAge <= N Rows( ages ), iAge += 1, // for each age...

	dtBigClass << selectwhere( age == ages[iAge] );
	dtSubset = dtBigClass << subset( selectedrows( 1 ), linked( 1 ), "invisible" ); 

	// the summary will be a single row with the requested information
	dtSummary = dtSubset << Summary( Mean( :height ), Mean( :weight ), Std Dev( :height ), Std Dev( :weight ), Freq( "None" ), Weight( "None" ) );
	// grab the values of interest in simpler variable names
	heightMean = dtSummary:name( "Mean(height)" )[1]; // row one has the answers
	heightStdDev = dtSummary:name( "Std Dev(height)" )[1];
	weightMean = dtSummary:name( "Mean(weight)" )[1];
	weightStdDev = dtSummary:name( "Std Dev(weight)" )[1];
	Close( dtSummary, "nosave" ); // close the work table, no longer needed
	
	// make a graph, attach to a borderbox so it does not open in a window...
	bb = Border Box(
		dtSubset << Graph Builder(
			Size( 400, 300 ),
			Show Control Panel( 0 ),
			Show Legend( 0 ), // don't need a legend
			Variables( X( :height ), Y( :weight ) ),
			Elements( Points( X, Y, Legend( 5 ) ) ), 
			// add ref lines. You can get a script like this GraphBuilder script by
			// making a graph builder interactively, then using the red triangle->get script.
			// below, the numbers in the script have been replaced with the variables above.
			SendToReport(
				Dispatch(
					{},
					"height",
					ScaleBox,
					{Add Ref Line( heightMean, "Dotted", "Red", "", 2 ), Add Ref Line( heightMean + heightStdDev, "Dotted", "Blue", "", 2 ),
					Add Ref Line( heightMean - heightStdDev, "Dotted", "Blue", "", 2 ), 
						// force all axes to be the same...
					Min( 48.2828947368421 ), Max( 75.8776488919668 ), Inc( 5 ), Minor Ticks( 1 )}
				),
				Dispatch(
					{},
					"weight",
					ScaleBox,
					{Add Ref Line( weightMean, "Dotted", "Red", "", 2 ), Add Ref Line( weightMean + weightStdDev, "Dotted", "Blue", "", 2 ),
					Add Ref Line( weightMean - weightStdDev, "Dotted", "Blue", "", 2 ), 
						// force all axes to be the same...
					Min( 55.2100840336134 ), Max( 188.127074359155 ), Inc( 20 ), Minor Ticks( 0 )}
				),
				Dispatch( {}, "Graph Builder", FrameBox, {Marker Size( 6 ), Transparency( 1 )} ), // make the markers bigger
				Dispatch( {}, "graph title", TextEditBox, {Set Text( "Age = " || Char( ages[iAge] ) )} ), // add the age= title
				Dispatch( {}, "Graph Builder", OutlineBox, {Set Title( "" )} ) // suppress the Graph Builder title
			)
		)
	);
	grid << append( bb ); // add the graph to the grid
); // the iAge loop

dtBigClass << clearRowStates; // undo the final selection
New Window( "the report", V List Box( H Center Box( Text Box( "Big Report", <<setFontSize( 20 ) ) ), grid ) );

Reports grouped in a 3-wide gridReports grouped in a 3-wide grid

Craige
Highlighted
DS
DS
Level VI

Re: How to Plot graph with reference lines for each element of table summary

Hi @spinup,

 

  I think I understand what you want. But, I'm also a little confused: are you wanting to summarize or just graph the 9 separate bivariate graphs?

 

  Working off your original data table, you can easily create all 9 graphs if you do a bivariate fit, e.g.:

Bivariate( 
  Y( :Score ), X( :Name( "# try" ) ), 
  By( :Name, :Grade )
);

  This will generate your nine graphs that you had colored in your previous post.

  Then, what you'll probably have to do is use a For Loop(s) to generate the quantile lines (or shaded area).

  The script for the first graph would be something like:

Bivariate(
	Y( :Score ),
	X( :Name( "# try" ) ),
	Where( :Name == "name1" & :Grade == "A" ), // this would be selecting 
	SendToReport(
		Dispatch(
			{},
			"2",
			ScaleBox,
			{Min( 75 ), Max( 100 ), Inc( 5 ), Minor Ticks( 1 ), // this sets the appearance of the y-axis
			Add Ref Line( 83, "Solid", "Blue", "25th quantile", 1 ),
			Add Ref Line( 95, "Solid", "Blue", "75th quantile", 1 )}
		)
	)
);

  In fact, if you set up a list of the names NameList={"Name1", "Name2"...}, or pull it from the column elements, then you can have a very short script by putting in For Loops and just us reference the NameList[i] element. This would be also the case for the Where(:Name==NameList[i] & Grade == GradeList[i]).

 

  Another option would be to inspect the tree structure of the output from the first script I included above and then you'd reference the AxisBox(i) for each of the y-axes for the 9 graphs. The y-axes are all odd, so if you do a For Loop, then you'd have to do i++ so that it iterates from 1 to 3 to 5 and so on in the loop.

 

  There's some really good stuff in the link that @Craige_Hales provided.

 

  For your specific issue, I think you'll definitely need to do some looping. Either that, or you can do the whole thing manually once and then just save the script to the data table (probably as a group), and then you can copy/paste that and save as a separate JSL file in a new script window. If you can figure out how to do it manually, JMP will write the JSL for you, and all you need to do is put it all together into a master JSL file that takes care of it all for you. The only issue there is if you decide to change names or column references or things like that. If you do, then you'll need to start using local variables.

 

Hope this helps!,

DS

View solution in original post

Article Labels

    There are no labels assigned to this post.