Subscribe Bookmark RSS Feed

How to export values behind the P(warning) charts in Shift Detection Profiler

Highlighted
scottahindle

Occasional Contributor

Joined:

May 18, 2017

In attached jmp file I have set up with some dummy data the shift detection profiler to have essentially zero measurement error and have a mean part shift of '3' equivalent to a 3-standard deviation process shift under the general assumption of normality with a mean of zero and an SD of 1.
When the jmp file is opened it simulates the situation for individual values where we want to detect a mean shift of 3 wtih the number of subgroups k equal to 10 (n=1).
I have attached two picture files to give an idea of what I want to do. First is with detection rule one only. Second is with detection rules 1 and 2 in the list (Wheeler Rule '4' choice of 8 points either side of the central line).
Is it possible to export the x-axis and y-axis values that are behind these great charts for the "Mean Shift" graph. Obviously, they are in the background so how to access/read them?
By exporting values for different scenarios I coud graphically compare the power functions and false alarm risks given the different choice of detection rules. (Here the emphasis is having these different scenarios on the same graph.)
Can anybody help? How to export and access the computed values for the x-axis and y-axis behind the graphs? (Main request is for the "mean shift" chart.)
Thanks as always for any assistance! Hope my points are clear.
Scott.

3 REPLIES
thomasz

Community Trekker

Joined:

Mar 11, 2016

Hi Scott

Here are two ways of getting the values from the axes:

pf=EMP Measurement Systems Analysis(
	Y( :Values ),
	X,
	Part( :Column 1 ),
	Model( "Crossed" ),
	Dispersion Chart Type( "Range" ),
	Parallelism Plots( 0 ),
	EMP Results( 1 ),
	Variance Components( 1 ),
	EMP Gauge RR Results( 1 ),
	Average Chart( 1 ),
	Dispersion Chart( 1 ),
	Shift Detection Profiler(
		Profiler(
			1,
			Confidence Intervals( 1 ),
			Term Value(
				Number of Subgroups( "10", Lock( 0 ), Show( 1 ) ),
				Part Mean Shift( 3, Lock( 0 ), Show( 1 ) ),
				Part Std Dev( 1, Lock( 0 ), Show( 1 ) ),
				Bias Factors Std Dev( 0, Max( 1 ), Lock( 0 ), Show( 1 ) ),
				Name( "Test-Retest Std Dev" )(0.0000008619, Lock( 0 ), Show( 1 )),
				Subgroup Size( "1", Lock( 0 ), Show( 1 ) )
			),
			Name( "Change In-Control Part Std Dev" )(1)
		)
	),
	Max Iter( 100 ),
	Conv Limit( 0.00000001 ),
	SendToReport(
		Dispatch(
			{},
			"Shift Detection Profiler",
			OutlineBox,
			{Set Summary Behavior( "Collapse" )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox,
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox( 3 ),
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox( 5 ),
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox( 7 ),
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox( 9 ),
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"EMP Shift Detection Profiler",
			FrameBox( 11 ),
			{Frame Size( 236, 214 )}
		),
		Dispatch(
			{"Shift Detection Profiler"},
			"Customize and Select Tests",
			OutlineBox,
			{Close( 0 )}
		)
	)
);

rep=pf<<report;
//Approach with xpath
ns=rep<<xpath("//OutlineBox[text()='Shift Detection Profiler']//AxisBox");
yPartMeanShiftMin=ns[1]<<get min;
yPartMeanShiftMax=ns[1]<<get max;
xPartMeanShiftMin=ns[4]<<get min;
xPartMeanShiftMax=ns[4]<<get max;

//2nd approach
yPartMeanShiftMin=rep(OutlineBox["Shift Detection Profiler"])[AxisBox(5)]<<get min;
yPartMeanShiftMax=rep(OutlineBox["Shift Detection Profiler"])[AxisBox(5)]<<get max;
xPartMeanShiftMin=rep(OutlineBox["Shift Detection Profiler"])[AxisBox(7)]<<get min;
yPartMeanShiftMax=rep(OutlineBox["Shift Detection Profiler"])[AxisBox(7)]<<get max;

KR, Thomas

scottahindle

Occasional Contributor

Joined:

May 18, 2017

Dear Thomas,

I have run the script a few times and while I always find the Measurement Systems Analysis output, including the Shift Detection Profiler output (the various graphs based on Number of Subgroups, Part Mean Shift...) I don't find, or know how to find, the actual values that are behind the graphs (meaning a series of x- and y-coordinates from which I could then create the same graphs outside of the Measurement Systems Analysis output forms).

If you could point out what I'm doing wrong, or not doing, please let me know. Would be much appreciated.

Regards, Scott.

thomasz

Community Trekker

Joined:

Mar 11, 2016

Dear Scott

 

I guess, now I understand what you are looking for. In some platforms you can simply save the profiler formula or the data grid, but not in this platform. A way could possibly be to change each coordinate sequentially and record the probability of Warning value for each, but as there a 6 dimensions, a grid with just 10 values will give you 10^6 combinations, so it is a dull path to go, and there is no guarantee for a usable result. Probably finding out the math behind and doing the calculations would be more feasible. Hopefully somebody comes up with a better solution.

 

Kind regards, Thomas