All diagnostics in the Design Diagnostics report are based on simulations that are computed from the sampling distributions of the appropriate mean squares. You can specify a simulation distribution and the parameter values for the distribution for each model factor. The design diagnostics update with changes to the settings. This enables you to explore the ability of your design to measure your systems under different assumed models. Update the number of levels in your design to explore alternatives.
The Variance Component Estimator Expected Performance report enables you to specify a simulation distribution and parameter values for each MSA model factor. These values are used in a large number of simulation trials to provide estimates of what you might observe in your study given your design and simulation assumptions. The buttons at the top of the report enable you to explore different models. Select between variance components or variance proportions estimates.
The Variance Component Estimator Expected Performance report contains model buttons and options for the simulation settings.
Use the model buttons to change the simulation model.
Full Model
Sets the simulation model to a full model.
Main Effects Only
Sets the simulation model to a main effects only model.
Interactions Up to Order
Sets the simulation model to a model with interactions up to the order specified.
Select between variance components or variance proportions simulated estimates.
Simulation Distribution
The simulation distribution for each model term. You can select between a fixed effect, random uniform, or random gamma distribution.
Distribution Parameters
The parameters for the specified distribution.
Variance Component
The name of the model term.
Estimate Plot
Plot of the estimate and the range of the simulated residuals. Green intervals indicate that the term is included in the model. Red intervals indicate a term that is not in the model.
Tip: Click the intervals to include or exclude terms from the model. You can also use the model buttons to change models.
Relative Bias
The average of the standardized residuals across the simulation trials. Each simulation has a true and an estimated variance component. The bias for each trial is the residual or difference between the estimated and true value. The residual divided by the actual value is the standardized residual.
Mean Absolute Relative Error
The average of the absolute value of the standardized residuals observed across the simulation trials. A value less than one is desired. A value less than one indicates that the level of error is less than the assumed true value.
The Variance Component Estimator Expected Performance red triangle menu contains the following option:
Make Data Table
Opens a data table of the simulation trials.The table contains the simulated true value and the estimated value for each trial.
A Gauge R&R report based on the simulated trials. Use to estimate the accuracy and sampling uncertainty that you can expect in these particular metrics based on the particular design.
An EMP report based on the simulated trials. For more information about EMP classification see“Monitor Classification Legend” in Quality and Process Methods.