The Design Space Profiler red triangle menu contains the following options:
Show Portion for Each Response
Shows or hides a column in the table of responses that contains the in-specification portion for each response at the current lower and upper limits for the factors. If there is more than one response, this column is shown by default.
Send Midpoints to Profiler
Sends the midpoints to the Prediction Profiler and sets the factors to the midpoint values. These values are the midpoints between the current Design Space Profiler lower and upper limits for the factors.
Get Midpoints from Profiler
Obtains the current factor settings from the Prediction Profiler and sets the midpoints for each factor in the Design Space Profiler to those values. Limits are constructed around each midpoint value using a specified fraction of the factor range.
Send Limits to Simulator
Sends the current limits for the continuous factors and a specified distribution to the Simulator. You can select a normal distribution with the limits at 2σ or 3σ, a normal weighted distribution with the limits at 2σ or 3σ, or a uniform distribution. The Simulator uses the Design Space Profiler factor limits and the specified distribution to define the parameters for the random factor distribution. If you select a normal or normal weighted distribution, the factor limits are used to determine the mean and the standard deviation, σ. If you select a uniform distribution, the factor limits are used to determine the lower and upper values of the distribution. This option also sends the Error StdDev value for each response to the Simulator. The values are used as the standard deviation of the added random noise for the corresponding response.
If there are categorical factors, this option sends an equal probability multinomial random distribution to the simulator. For factor levels that have been excluded in the Design Space Profiler, their probabilities are set to zero in the Simulator.
Note: This option launches the Simulator if it has not been launched yet.
Tip: Use the weighted normal distributions when defects are a very rare event and you want the simulation sampler to explore the space in the tails of the distribution.
Send Limits to Profiler as Constraints
Sends the current factor limits in the Design Space Profiler to the Prediction Profiler as bound constraints. If the Simulator is turned on and is set to Uniform, the simulator bounds are adjusted to be feasible with the profiler bound constraints.
Note: To remember the profiler bounds, save the constraints as a data table script using the Save Linear Constraints menu option.
Reset Factor Space
Opens a window that enables you to change the factor space by narrowing, widening, or shifting the range of one or more factors. The window shows the current minimum, maximum, lower limit, and upper limit for each factor. You can enter new values for the minimum and maximum of each factor individually or you can use the Make Narrower button to set to narrow the ranges of all factors, centered across the limits.
Make and Connect Random Table
Creates a new data table that contains uniformly distributed factor settings and their corresponding simulated responses. The new data table is connected to the design space profiler with the rows that contain factors that are within the current lower and upper limits selected. When you select this option, a window is shown enabling you to specify how the responses are simulated. You can specify the number of runs, specify the use of random noise, and update the response standard deviation values. You can also specify to save confidence or prediction intervals to the data table and to embed the factor or response space plots to your report.
Scatterplots of the factor space and the response space provide a visual representation of the operating limits and their impact on the in-spec results. The response space plots are available only for profilers with two or more responses. View the plots by selecting the embed option or by running the corresponding scripts in the new data tables. Predicted responses that are within the specification limits are colored green and predicted responses that are outside of the specification limits are colored red.
Connect to Table
Connects the Design Space Profiler report to the specified data table. The rows that contain factors that are within the current lower and upper limits are selected in the connected table. The row selections are automatically updated as changes are made in the Design Space Profiler report.
Connect Hide Mode
(Available only after the Connect to Table option is selected.) Hides the rows in the connected data table that contain factors that are outside of the current lower and upper limits.
Show Current Profiler Values
Shows or hides the current values from the Prediction Profiler as vertical gray dotted reference lines.
Save X Spec Limits
Saves the current operating limits for the factors as spec limits column properties in the original data table.
Show Corners
Shows or hides the Corners report. This report contains a table of the in-specification probabilities at the extremes of the factor space. See Corners Report. When this option is selected, the corner with the worst in-specification probability is added to the table in the Design Space Profiler report that contains the overall in-spec portion and volume portion.
The Corners report provides information about the in-specification probabilities at the extremes of the factor space. The extremes of the factor space are denoted as corners and represent factor combinations at the high and low limits of each factor. Categorical factors are included and consider all non-excluded levels as corners. If the limits are updated in the Design Space Profiler, the Corners report is updated automatically.
The Corners report contains a table of in-specification probabilities. The table contains the following columns:
Factor Corner
The factor combination for each corner. A “+” represents the upper limit of a factor and a “-” represents the lower limit of a factor.
InSpec Prob
The in-spec probability for each factor combination. There is a probability column for each response variable. The probabilities are calculated using a normal distribution centered at the predicted values and cut at the specification limits. If no standard deviation is specified, all of the probabilities are zero or one.
Overall
(Available only if there are multiple responses.) The overall in-spec probability. This is calculated by assuming that the responses are independent and multiplying the individual probabilities for each response.
-log10 Prob
A calculated probability value that is used to move the corners when the in-spec probabilities are very low. Large values indicate that the probability of being in-spec is very low. The -log10(prob) value is calculated as follows:
– If there is only one response column with only one specification limit, the value is -log10(prob), where prob is the in-spec probability.
– If there is only one response column with both lower and upper specification limits, the value is -log10(prob), where prob is the smaller in-spec probability.
– If there are multiple responses, the value is -log10(prob1) +... + -log10(probr), where prob1,..., probr are the one-sided in-spec probabilities for the responses.
Move Corner Inward, Move Corner Outward
Use the Move Corner Inward and Move Corner Outward buttons to explore how adjusting the factor extremes changes the in-spec probabilities. The Move Corner Inward button finds the factor limit that improves the Worst Corner in-spec probability the most and moves that factor limit inward. These changes are tracked in a history table at the bottom of the report.
Note: The Move Corner Inward and Move Corner Outward buttons are not available when there are categorical factors.
If all factors are continuous and there are between two and six factors, the probabilities are also shown in cube plots. For more information about cube plots, see “Cube Plots” in Fitting Linear Models.