Introduction to Profiling

It is easy to visualize a response surface with one input factor X and one output factor Y. It becomes harder as more factors and responses are added. The profilers in JMP provide a number of highly interactive cross-sectional views of any response surface.

Desirability profiling and optimization features are available to help find good factor settings and produce desirable responses. Most profilers also incorporate multithreading for faster computation. Simulation and defect profiling features are available for when you need to make responses that are robust and high-quality when the factors have variation.

Profiling Features in JMP

There are several profiler facilities in JMP, accessible from a number of fitting platforms and the main menu under Graph. They are used to profile data column formulas.

Profiler Features Summary
	Description	Features
Profiler	Shows vertical slices across each factor, holding other factors at current values	Desirability, Optimization, Simulator, Propagation of Error
Contour Profiler	Horizontal slices show contour lines for two factors at a time	Simulator
Surface Profiler	3-D plots of responses for 2 factors at a time, or a contour surface plot for 3 factors at a time	Surface Visualization
Mixture Profiler	A contour profiler for mixture factors	Ternary Plot and Contours
Custom Profiler	A non-graphical profiler and numerical optimizer	General Optimization, Simulator
Excel Profiler	Visualize models (or formulas) stored in Excel worksheets.	Profile using Excel Models

Profiler availability is shown in Where to Find JMP Profilers.

Where to Find JMP Profilers
Location	Profiler	Contour Profiler	Surface Profiler	Mixture Profiler	Custom Profiler
Graph Menu (as a Platform)	Yes	Yes	Yes	Yes	Yes
Fit Model: Least Squares	Yes	Yes	Yes	Yes
Fit Model: Logistic	Yes
Fit Model: LogVariance	Yes	Yes	Yes
Fit Model: Generalized Linear	Yes	Yes	Yes
Nonlinear: Factors and Response	Yes	Yes	Yes
Nonlinear: Parameters and SSE	Yes	Yes	Yes
Generalized Regression	Yes
Mixed Model	Yes	Yes	Yes	Yes
Neural Net	Yes	Yes	Yes
Gaussian Process	Yes	Yes	Yes
Custom Design Prediction Variance	Yes		Yes
Life Distribution	Yes
Fit Life by X	Yes		Yes
Choice	Yes

Note: In this guide, we use the following terms interchangeably:

•	factor, input variable, X column, independent variable, setting

•	response, output variable, Y column, dependent variable, outcome

The Profiler (with a capital P) is one of several profilers (lowercase p). Sometimes, to distinguish the Profiler from other profilers, we call it the Prediction Profiler.

When the profiler is invoked as a platform from the main menu, rather than through a fitting platform, you provide columns with formulas as the Y, Prediction Formula columns. These formulas could have been saved from the fitting platforms.

Profiler Launch Window

The columns referenced in the formulas become the X columns (unless the column is also a Y).

Y, Prediction Formula

The response columns containing formulas.

Noise Factors

Only used in special cases for modeling derivatives. Details are in the Noise Factors section.

Expand Intermediate Formulas

Tells JMP that if an ingredient column to a formula is a column that itself has a formula, to substitute the inner formula, as long as it refers to other columns. To prevent an ingredient column from expanding, add an Other column property, name it Expand Formula, and assign a value of 0.

The Surface Plot platform is discussed in a separate chapter. The Surface Profiler is very similar to the Surface Plot platform, except Surface Plot has more modes of operation. Neither the Surface Plot platform nor the Surface Profiler have some of the capabilities common to other profilers.

Fit Group

For the REML and Stepwise personalities of the Fit Model platform, if models are fit to multiple Y’s, the results are combined into a Fit Group report. This enables the different Y’s to be profiled in the same Profiler. The Fit Group red triangle menu has options for launching the joint Profiler. Profilers for the individual Y’s can still be used in the respective Fit Model reports.

Fit Group reports are also created when a By variable is specified for a Stepwise analysis. This allows for the separate models to be profiled in the same Profiler.

The Fit Group scripting command can be used to fit models in different platforms, and have the individual models profiled in the Profiler. For more details, see the Scripting Guide.