Model Profiler Introductory Overview
The Model Profiler makes it possible for you to run simulations based on a specified model in the Experimental Design (DOE), General Linear Models (GLM), and Distributions & Simulation modules.
The Model Profiler is used similarly in the DOE and GLM modules. Specifically, a model is specified within the respective module. After the model is specified, the Model Profiler is activated by clicking the Model Profiler button. A simulation is completed using a number of runs, 5,000 runs by default, or any number of runs you specify.
In the Distributions & Simulation module, instead of specifying a model, you can load models that have already been specified and saved in PMML files. After running simulations on a series of models, the results of those analyses can then be compared.
While the Model Profiler can be accessed through three different modules, the interface is the same in each module, except that the Load models button is not available when the Model Profiler is accessed from the DOE and GLM modules.
The Model Profiler contains two tabs: Profiler and Output.
The Profiler tab lists the predictor variables that have been specified for the model. When you click on the predictor variable, you can specify the appropriate distribution for that variable as well as the properties for that distribution. If the distribution is not within the list of available distributions, you can add your own user-specified distribution. If you prefer, you can opt for STATISTICA to fit the best distribution to each of the predictor variables. Finally, you can specify the correlations between the predictor variables via the Edit Correlation button.
On the Output tab, you can set the specifications for the simulation. You can specify the number of runs for the simulation, edit specifications for each of the predictor variables, and obtain the predictor formula for the model. By default, STATISTICA sets the number of runs at 5,000, but this can be customized. Each predictor variable can be selected in a drop-down list. After specifying the predictor variables, noise can be added using a standard deviation that you can set. Moreover, design specifications can be set specific to the predictor variables including a lower specification limit, upper specification limit, and target value. After these have been set, you can review capability specifications, summary statistics, and summary graphs for each of the predictor variables. Finally, the prediction equation is generated in the predictor formula box.
In addition to the output generated through the Capability Stats, Summary Stats, and Summary Graph buttons, data files are automatically generated via the Model Profiler. One data file generated includes the runs specified. If lower and upper specification limits and a target value are specified, a data file with defect rates is generated.