Adding and Viewing Column Properties
The DOE platforms automatically save certain column properties to the design tables that they construct. However, some of the column properties associated with designed experiments are useful in general modeling situations. To use these column properties with data tables that have not been created using DOE platforms, you can add them yourself.
Adding a Column Property
To assign a column property to one or more columns, do the following:
1. Select the column or columns to which you want to assign a property.
2. Do one of the following:
‒ Right-click the header area, select Column Properties, and select the property.
‒ Right-click the header area, select Column Info, and select the property from the Column Properties menu.
‒ Select Cols > Column Info and select the property from the Column Properties menu.
3. In the column property panel that appears, specify values and select options as appropriate.
4. Click Apply to add the column property or click OK to add the column property and close the column properties window.
Tip: A column might already contain a property that you want to apply to other columns. Use the Standardize Attributes command to apply that property to other columns. For details, see The Column Info Window chapter in the Using JMP book.
Viewing a Column Property
To view the properties assigned to a specific column, in the columns panel, click the column property asterisk icon 
. Click a property to see its settings or to edit it. Figure A.2 shows the column properties assigned to Stretch in the Bounce Data.jmp sample data table, located in the Design Experiment folder.
. Click a property to see its settings or to edit it. Figure A.2 shows the column properties assigned to Stretch in the Bounce Data.jmp sample data table, located in the Design Experiment folder.Figure A.2 Asterisk Icon for Stretch Revealing Two Column Properties
You can also view column properties by accessing the Column Properties list in the Column Info menu. Select the column or columns whose column properties you want to view and do one of the following:
• Right-click the header area, select Column Info, and select the property from the Column Properties list.
• Select Cols > Column Info and select the property from the Column Properties list.
Response Limits
Using the Response Limits column property, you can specify the following:
• bounds on the range of variation for a response
• a desirability goal
• a measure of the importance of the response
• desirability values
The Response Limits column property defines a desirability function for the response. The Profiler and Contour Profiler use desirability functions to find optimal factor settings. See the Profiler chapter in the Profilers book.
Figure A.3 shows the Response Limits panel in the Column Info window for the response Stretch in the Bounce Data.jmp sample data table, found in the Design Experiment folder.
Figure A.3 Example of the Response Limits Panel
The Response Limits panel consists of the following areas:
Goal
Select your response goal from the menu. Available goals are Maximize, Match Target, Minimize, and None. JMP defines a desirability function for the response to match the selected goal. If you specify limits, the desirability function is defined using these limits. If you do not specify limits, JMP bases the desirability function on conservative limit values derived from the distribution of the response. If None is selected as the goal, then all response values are considered equally desirable. For further details, see “Responses” in the “Custom Designs” chapter.
Importance
Enter a relative weighting for each of several responses in computing the overall desirability function. The Importance value can be any positive number. When no Importance value is specified, JMP treats all responses in a given analysis as having equal Importance values. If there is only one response, it receives Importance 1.
Value
Specify Lower and Upper limits and a Middle value for your response. JMP uses these values to construct a desirability function for the response. If you do not specify limits, JMP bases the desirability function on conservative limit values. If your goal is Match Target and no Middle value is specified, then the target is defined to be the midpoint of the Lower and Upper limits.
Desirability
Specify values that reflect the desirability of your Lower, Middle, and Upper values. Desirability values should be between 0 and 1. If you do not specify Desirability values, JMP assigns values in accordance with the selected Goal.
Show as graph reference lines
Shows horizontal reference lines for the Lower, Middle, and Upper values in the Actual by Predicted Plot and the Prediction Profiler. This option applies only if limits are specified.
Response Limits Example
The Coffee Data.jmp sample data table (located in the Design Experiment folder) contains the results of an experiment that was performed to optimize the Strength of coffee. For a complete description of the experimental design and analysis, see “The Coffee Strength Experiment” in the “Starting Out with DOE” chapter.
Your goal is to find factor settings that enable you to brew coffee with a target strength of 1.3, which is considered to be the most desirable value. Values less than 1.2 and greater than 1.4 are completely undesirable. The desirability of values between 1.2 and 1.4 decreases as their distance from 1.3 increases.
1. Select Help > Sample Data Library and open Design Experiment/Coffee Data.jmp.
2. In the Table panel, click the green triangle next to the DOE Dialog script.
The DOE Dialog script re-creates the Custom Design dialog that was used to create the experimental design in Coffee Data.jmp.
3. Open the Responses outline.
Figure A.4 Responses Outline in Custom Design Window
When you designed this experiment, you specified a response Goal of Match Target with a Lower Limit of 1.2 and an Upper Limit of 1.4. Since there is only one response, you did not specify a value for Importance, because it is 1 by default. When you constructed the design table, JMP assigned the Response Limits column property to Strength.
4. Close the Custom Design window.
5. In the Coffee Data.jmp sample data table, select the Strength column and select Cols > Column Info.
6. Select Response Limits in the Column Properties list.
Figure A.5 Response Limits Column Property for Strength
Notice the following:
‒ The Goal is set to Match Target.
‒ Importance is missing. When Importance is missing, JMP treats all responses in a given analysis as having equal Importance values. So JMP assigns Strength an Importance value of one.
‒ The Lower limit is 1.2.
‒ The Upper limit is 1.4.
‒ No Middle value is specified.
Because no Middle value is specified, the target is defined to be the midpoint of the Lower and Upper limits, which is 1.3.
‒ No Desirability values are specified.
7. Select the Show as graph reference lines option.
This option shows horizontal reference lines for the Lower, Middle, and Upper values in the Actual by Predicted Plot and the Prediction Profiler.
8. Click OK.
9. In the Coffee Data.jmp data table, click the green triangle next to the Reduced Model script.
10. Click Run.
The Prediction Profiler appears at the bottom of the report.
Figure A.6 Profiler Showing Desirability Function for Strength
The desirability function for Strength appears in the plot at the right above Desirability. This plot appears because the data table contains a Response Limits column property for Strength. The Prediction Profiler also shows reference lines for the Lower and Upper limits for Strength.
11. Hold down the Ctrl key and click the Strength plot for Desirability.
Figure A.7 Response Goal Window for Strength
Notice the following:
‒ JMP determines the Middle value to be the midpoint of the High and Low limits that you specified in the Response Limits column property.
‒ Because the Goal is set to Match Target, JMP sets the Desirability for the Middle value to 1.
‒ JMP sets the Desirability for the High and Low values to very small numbers, 0.0183.
‒ The Desirability plot in Figure A.6 shows how JMP uses the Desirability values shown in Figure A.7. The Desirability function for Strength is essentially 0 beyond the Low and High values and it increases to 1 gradually as Strength approaches the target of 1.3. The Importance value is set to 1 since there is only one response in the model.
12. Click Cancel to exit the window.
13. Select Optimization and Desirability > Maximize Desirability from the Prediction Profiler red triangle menu.
The settings for Time and Charge are updated to show settings for the factors that maximize the desirability function for Strength. However, many other settings also maximize the desirability function. See the Contour Profiler chapter in the Profilers book for information about how to identify other settings that maximize the desirability function.
Editing Response Limits
In the Vinyl Data.jmp sample data table, a Response Limits column property is already assigned to the response thickness. The property has a goal of maximizing thickness. Suppose that instead of maximizing thickness, you want the sheets of vinyl to have a thickness between 6 and 10, with a target thickness of 8.5.
1. Select Help > Sample Data Library and open Design Experiment/Vinyl Data.jmp.
2. Select the thickness column and select Cols > Column Info.
The Response Limits property appears in the Column Properties list as the only property assigned to thickness. The Response Limits panel appears to the right of the list.
3. Click Maximize and select Match Target.
4. Type 1 for the Importance value.
5. Under Value, type 6 for the Lower value, 8.5 for the Middle value, and 10 for the Upper value.
This is an example of asymmetric response limits. Values of thickness as small as 6 or as large as 10 are acceptable. However, the target for thickness is 8.5.
6. Select Show as graph reference lines.
This option shows horizontal reference lines for the Lower, Middle, and Upper values in the Actual by Predicted Plot and the Prediction Profiler.
Figure A.8 Completed Response Limits Panel
7. Click OK.
8. In the Vinyl Data.jmp data table, click the green triangle next to the Model script.
Note that m1, m2, and m3 are mixture factors. Also, the design involves a random Whole Plots factor. Because of this, the default Method is REML (Recommended).
9. Click Run.
10. From the red triangle next to Response thickness, select Row Diagnostics > Plot Actual by Predicted.
The reference lines for the Lower, Middle, and Upper limits appear on the Actual by Predicted Plot.
11. From the red triangle next to Response thickness, select Factor Profiling > Profiler.
Figure A.9 Prediction Profiler Showing Asymmetric Desirability Function
The plot at the right above Desirability shows the desirability function that JMP has constructed for thickness. The desirability is 1 at the Middle limit of 8.5. The desirability is essentially 0 for thickness values below 6 and above 10.
12. Hold down the Ctrl key and click the thickness plot for Desirability.
Figure A.10 Response Goal Window for Thickness
This window shows your settings for the High, Middle, and Low Values. It also shows the Desirability values that JMP assigns, based on your goal of Match Target. The Desirability function shown in Figure A.9 is a continuous curve that matches the Desirability settings in Figure A.10 at the High, Middle, and Low Values. At other values, the Desirability function assigns desirabilities that are consistent with the selected goal.
13. Click Cancel.
14. Select Optimization and Desirability > Maximize Desirability from the Prediction Profiler red triangle menu.
The settings for the factors are updated to show values that maximize the desirability function for thickness. Keep in mind that many other settings also maximize the desirability function. The predicted response at these optimal settings is 8.5. Recall that you set 8.5 as the target setting, with limits of 6 and 10.
15. Close the Vinyl Data.jmp sample data table without saving the changes.
Design Role
Factors in designed experiments, as well as in more general models, can behave in various ways. JMP uses the design role column property to describe these behaviors. These are the possible design roles:
• Continuous
• Discrete Numeric
• Categorical
• Blocking
• Covariate
• Mixture
• Constant
• Uncontrolled
• Random Block
• Signal
• Noise
In many of the JMP DOE platforms, you can specify factors with different design roles. In some platforms, your design requirements cause JMP to define factors. For example, Whole Plots and Subplots are factors that JMP creates when you specify very-hard-to-change and hard-to-change factors. In platforms where various design roles can occur, when JMP creates the design table for your design, each factor is assigned the Design Role column property.
• For descriptions of the design roles other than Random Block, Signal, and Noise, see “Factor Types” in the “Custom Designs” chapter.
• For a description of the Random Block design role, see “Changes and Random Blocks” in the “Custom Designs” chapter.
• For descriptions of the Signal and Noise design roles, see “Factors” in the “Taguchi Designs” chapter.
Design Role Example
The experiment in the Odor Control Original.jmp sample data table studies the effect of three factors on odor. You designed the 15-run experiment and it was conducted. However, when the results were reported to you, you learned that the experiment was conducted over three days. The first 5 runs were conducted on Day 1, the second 5 runs on Day 2, and the remaining 5 runs on Day 3.
Since variations in temperature and humidity might have an effect on the response, you want to include Day as a random blocking factor. It is easy to add a column for Day to the design table. But you want to use the Evaluate Design platform to compare the design with the unexpected block to your original design. You also want the ability to use the Augment Design platform in case you need to augment the design. To use the Evaluate Design and Augment Design platforms, you need to add the Design Role column property to your new Day column.
1. Select Help > Sample Data Library and open Odor Control Original.jmp.
2. Select the first column, Run.
3. Select Cols > New Columns.
4. Next to Column Name, type Day.
5. From the Initialize Data list, select Sequence Data.
6. Enter the following:
‒ 1 for From
‒ 3 for To
‒ 1 for Step
‒ 5 for Repeat each value N times
7. Next to Number of columns to add, type 1.
‒ Click OK.
Figure A.11 Completed New Columns Window
8. Click OK.
The Day column is added as the second column in the data table.
9. Select the Day column.
10. Select Cols > Column Info.
11. From the Column Properties list, select Design Role.
12. In the Design Role panel, click Continuous and select Random Block.
13. Click OK.
In the columns panel, an asterisk appears next to Day.
14. Click the asterisk next to Day to verify that the Design Role column property has been assigned.
15. Close the Odor Control Original.jmp sample data table without saving the changes.
Coding
The Coding column property applies only to columns with a numeric data type. It applies a linear transformation to the data in the column. In the Coding column property window, you specify a Low Value and a High Value. The Low Value and High Value in your original data are transformed to –1 and +1. JMP uses the transformed data values whenever the column is entered as a model effect in the Fit Model platform.
The coding property is useful for the following reasons:
• Coded predictors lead to parameter estimates that are more easily interpreted and compared.
• Coded predictors help reduce multicollinearity in models with interaction and higher-order terms.
When any DOE platform other than Accelerated Life Test Design creates a design, JMP defines a Coding column property for each non-mixture factor with a numeric data type. Figure A.12 shows the Coding column property panel for the column Feed Rate in the Reactor 20 Custom.jmp sample data table, found in the Design Experiment folder.
Figure A.12 Coding Property Panel for Feed Rate
Low and High Values
When the Coding property is applied as part of design construction in a DOE platform, the Low Value is initially set to the minimum setting of the factor and the High Value is initially set to the maximum setting.
When you apply the Coding property to a column that does not contain that property, JMP inserts the minimum value as the Low Value and the maximum value as the High Value. You can change these values as needed.
Caution: After the Coding column property is assigned to a column, JMP does not automatically update it when you make changes to the values in the column. If you change the values in a column that has a Coding column property, review the High Value and Low Value to ensure that they are still appropriate.
The Coding column property centers each value in a column by subtracting the midpoint of the High Value and Low Value. It then divides by half the range. Suppose that H is the High Value and L is the Low Value. Then every X in the column is transformed to the following:
For each factor, the transformed values have a midpoint equal to 0 and range from -1 to +1.
Coding Column Property and Center Polynomials
The Center Polynomials option is located in the Fit Model launch window, within the Model Specification red triangle menu. Center Polynomials centers a continuous column involved in a polynomial term by subtracting the mean of each value in the column. For more details, see the Model Specification chapter in the Fitting Linear Models book.
If the Coding column property is assigned to a column, then the Center Polynomials option has no effect on that column. In a polynomial term involving that column, the values are centered and scaled as specified by their Coding property. Suppose that another column in the model does not have the Coding property and that you select Center Polynomials. Then that column is centered by its mean in any polynomial term where it appears.
Coding Example
The Reactor 20 Custom.jmp sample data table contains data from a 20-run design that was constructed using the Custom Design platform. The experiment investigates the effects of five factors on a yield response (Percent Reacted) for a chemical process.
1. Select Help > Sample Data Library and open Design Experiment/Reactor 20 Custom.jmp.
2. In the Table panel, click the green triangle next to the DOE Dialog script.
3. Open the Factors outline.
Figure A.13 Factors Outline for Design Used in Reactor 20 Custom.jmp
Notice that the settings for Temperature range from 140 to 180. When the design was generated, the Coding column property was assigned to Temperature. The Low Value is set to 140 and the High Value is set to 180.
4. Close the Custom Design window.
5. In the Reactor 20 Custom.jmp sample data table, click the asterisk next to Temperature in the columns panel and select Coding.
The Column Info window appears and shows the Coding column property panel. You can see that JMP added the column property, specifying the Low Value and High Value, when it constructed the design table. In fact, by repeating this step, you can verify that JMP added the Coding property for all five factors.
Figure A.14 Coding Panel for Temperature
6. Click Cancel to close the Column Info window.
7. In the Reactor 20 Custom.jmp sample data table, click the green triangle next to the Reduced Model script.
This script fits a model that contains only the five effects determined to be significant based on an analysis of the full model.
8. Click Run.
Figure A.15 Effect Summary Report for Reduced Model
In the Source list, the High and Low values used in the Coding column property appear in parentheses to the right of the main effects, Catalyst, Temperature, and Concentration. The ranges imposed by the Coding property are not shown for the interaction effects.
Tip: Notice the “^” symbols to the right of the PValues for Temperature and Concentration. These symbols indicate that these main effects are components of interaction effects with smaller p-values. If an interaction effect is included in the model, then the principle of effect heredity requires that all component effects are also in the model. See “Effect Heredity” in the “Starting Out with DOE” chapter.
9. Select Estimates > Show Prediction Expression from the red triangle next to Response Percent Reacted.
Look at the Prediction Expression outline to see how coding affects the prediction formula.
Figure A.16 Prediction Expression for Reduced Model
Each factor is transformed as specified by the Coding column property. For example, for Temperature, notice the following:
‒ The Low Value in the Coding property was set to 140. The Temperature value of 140 is transformed to -1.
‒ The High Value in the Coding property was set to 180. The Temperature value of 180 is transformed to +1.
‒ The midpoint of the Low and High values is 160. The Temperature value of 160 is transformed to 0.
The transformed values help you compare the effects. The estimated coefficient for Catalyst is 9.942 and the estimated coefficient for Concentration is -3.077. It follows that the predicted effect of Catalyst on Percent Reacted is more than three times as large as the effect of Concentration on Percent Reacted. Also, the coefficients indicate that predicted Percent Reacted increases as Catalyst increases and decreases as Concentration increases.
The transformed values help you interpret the coefficients:
• When all factors are at their midpoints, their transformed values are 0. The predicted Percent Reacted is the intercept, which is 65.465.
• When Catalyst and Concentration are at their midpoints, a 20 unit increase in Temperature increases the Percent Reacted by 5.558 units.
• Suppose that Concentration is at its midpoint, so that its transformed value is 0:
‒ When Catalyst is at its midpoint, a 20 unit increase in Temperature increases the Percent Reacted by 5.558 units.
‒ When Catalyst is at its high setting, a 20 unit increase in Temperature increases the Percent Reacted by 5.558 + 6.035 = 11.593 units.
It follows that the coefficient of the interaction term, 6.035, is the increase in the slope of the model for predicted Percent Reacted for a 0.5 unit change in Catalyst.
Assigning Coding
The experimental data in the Tiretread.jmp sample data table results from an experiment to study the effects of SILICA, SILANE, and SULFUR on four measures of tire tread performance. In this example, you will consider only one of the responses, ABRASION.
You will first fit a model using the uncoded factors. Then you will assign the coding property to the factors and rerun the model to obtain meaningful parameter estimates.
1. Select Help > Sample Data Library and open Tiretread.jmp.
2. Select Analyze > Fit Model.
3. Select ABRASION and click Y.
4. Select SILICA, SILANE, and SULFUR and click Macros > Response Surface.
5. Check Keep dialog open.
6. Click Run.
7. Select Estimates > Show Prediction Expression from the Response ABRASION red triangle menu.
Figure A.17 Prediction Expression for Model with Uncoded Factors
The coefficients do not help you compare effect sizes. The sizes of the coefficients do not reflect the impact of the effects on ABRASION over the range of their settings. Also, the coefficients are not easily interpreted. For example, the coefficients do not facilitate your understanding of the predicted response when SILICA is at the midpoint of its range.
Apply the Coding column property to the three factors to see how coding makes the coefficients more meaningful.
8. In the Tiretread.jmp data table, select SILICA, SILANE, and SULFUR in the Columns panel. Right-click the highlighted column titles and select Standardize Attributes.
9. Select Column Properties > Coding in the Standardize Properties panel.
10. Click OK.
An asterisk appears in the Columns panel next to SILICA, SILANE, and SULFUR indicating that these have been assigned a column property.
11. In the Fit Model window, click Run.
12. Select Estimates > Show Prediction Expression from the Response ABRASION red triangle menu.
Figure A.18 Prediction Expression for Model with Coded Factors
The coefficients for the coded factors enable you to compare effect sizes. SILANE has the largest effect on ABRASION over the range of design settings. The effects of SILICA and the SILANE*SULFUR interaction are large as well.
The coefficients for the coded factors are also more easily interpreted. For example, when all factors are at the center of their ranges, the predicted value of ABRASION is the intercept, 139.12.
13. Close the Tiretread.jmp sample data table without saving the changes.
Mixture
The Mixture column property is useful when a column in a data table represents a component of a mixture. The components of a mixture are constrained to sum to a constant. Because of this, they differ from non-mixture factors. The Mixture column property serves two purposes:
• It identifies a column as a mixture component.
If you add a column with the Mixture column property as a model effect in the Analyze > Fit Model window, JMP automatically generates a no-intercept model.
• It defines the coding for a mixture component.
Coding for mixture components differs from that for non-mixture factors. However, as with non-mixture factors, a benefit of coding for mixture factors is that it helps you interpret parameter estimates. See “PseudoComponent Coding”.
Figure A.19 shows the Mixture column property panel for the factor m1 in the Vinyl Data.jmp sample data table, found in the Design Experiment folder.
Figure A.19 Mixture Column Property Panel
In the Mixture column property panel, you can specify the following:
Lower Limit
Specifies the low value used in PseudoComponent Coding. When the Mixture property is applied as part of design construction in a DOE platform, the Lower Limit is set to the minimum setting of the factor. When you apply the Mixture property to a column that does not contain that property, JMP inserts the minimum value as the Lower Limit. You can change this value as needed.
Upper Limit
Specifies the high value used in PseudoComponent Coding. When the Mixture property is applied as part of design construction in a DOE platform, the Upper Limit is set to the maximum setting of the factor. When you apply the Mixture property to a column that does not contain that property, JMP inserts the maximum value as the Upper Limit. You can change this value as needed.
Sum of Terms
Specifies the sum of the mixture components. When you apply the Mixture property to a column that does not contain that property, JMP inserts a default value of 1 for the Sum of Terms.
L PseudoComponent Coding
Transforms data values so that the Lower Limit corresponds to 0.
U PseudoComponent Coding
Transforms data values so that the Upper Limit corresponds to 0.
PseudoComponent Coding
A pseudo-component is a linear transformation. Let S denote the sum of the mixture components. Suppose that i columns 
have been assigned the Mixture column property. Suppose that the columns and effects constructed from these columns are entered as effects in the Fit Model window.
have been assigned the Mixture column property. Suppose that the columns and effects constructed from these columns are entered as effects in the Fit Model window.Define the following:
, where Li is the Lower Limit for Xi
, where Ui is the Upper Limit for XiLet xi denote a value of the column Xi. The L PseudoComponent at xi is defined as follows:
The U PseudoComponent at xi is defined as follows:
If you select both L PseudoComponent Coding and U PseudoComponent Coding, the Fit Model platform uses the L coding if (S – L) < (U – S). Otherwise, the U coding is used.
In Fit Model, mixture factors are transformed using pseudo-components before computing parameter estimates. This helps make parameter estimates more meaningful. In reports dealing with parameter estimates, the mixture main effects are given by the coding transformation. Other reports, such as the profilers, are based on the uncoded values.
Mixture Example
The data in the Donev Mixture Data.jmp sample data table, found in the Design Experiment folder, are based on an example from Atkinson and Donev (1992). The design includes three mixture factors and one non-mixture factor. The response and factors are as follows:
• The response is the electromagnetic Damping of an acrylonitrile powder.
• The three mixture ingredients are copper sulphate (CuSO4), sodium thiosulphate (Na2S2O3), and Glyoxal.
• The non-mixture environmental factor of interest is the Wavelength of light.
Though Wavelength is theoretically continuous, the researchers were interested only in predictions at three discrete wavelengths. As a result, Wavelength is treated as a categorical factor with three levels.
For details about using Custom Design to construct a design for this situation, see “Mixture Experiments” in the “Examples of Custom Designs” chapter.
1. Select Help > Sample Data Library and open Design Experiment/Donev Mixture Data.jmp.
2. Click the asterisk next to CuSO4 in the columns panel and select Mixture.
Figure A.20 Mixture Column Property Panel for CuSO4
Notice the following:
‒ The Lower Limit is 0.2, the minimum design setting for CuSO4.
‒ The Upper Limit is 0.8, the maximum design setting for CuSO4.
‒ The Sum of Terms is set to 1. This is the sum of the three mixture factors.
‒ The L PseudoComponent Coding option is selected. See “PseudoComponent Coding”.
3. Click Cancel.
4. Click the asterisk next to Glyoxal in the columns panel and select Mixture.
For this factor, note the following:
‒ The Lower Limit is 0, the minimum design setting for Glyoxal.
‒ The Upper Limit is 0.6, the maximum design setting for Glyoxal.
5. Click Cancel.
6. In the Donev Mixture Data.jmp data table, click the green triangle next to the Model script.
The model contains the main effects of the mixture factors and two-way interactions for all four factors.
7. Click Run.
In the Parameter Estimates report, the mixture factors appear in their pseudo-component coded form. When the mixture factors appear in interactions, they are not denoted in coded form. Nevertheless, the model fitting is based on the pseudo-components. The first three terms in the Parameter Estimates report (Figure A.21), show the coded form for the mixture factors.
Figure A.21 Parameter Estimates Report
8. Select Estimates > Show Prediction Expression from the Response Damping red triangle menu.
The Prediction Expression report shows the model that was fit. Note that the mixture factors are transformed using the L PseudoComponent coding.
Figure A.22 Prediction Expression for Damping Model
Suppose that you are interested in predictions at Wavelength L2. Suppose also that Na2S2O3 and Glyoxal are set to their low values, 0.2 and 0 respectively, and that CuSO4 is set to its high value, 0.8. In this case, the predicted Damping equals the parameter estimate for CuSO4 (6.191) plus the parameter estimate for CuSO4*Wavelength[L2] (1.878). You can verify this in the Prediction Profiler.
9. Select Save Columns > Save Coding Table from the Response Damping red triangle menu.
Figure A.23 First Three Columns of Coding Table Showing Coded Mixture Factors
For this particular design, the L PseudoComponent coding transforms the mixture factors to range between 0 and 1. Note that this does not happen in general.
Factor Changes
The Factor Changes column property indicates how difficult it is to change factor settings in a designed experiment. The possible specifications for Factor Changes are Easy, Hard, and Very Hard. For example, Figure A.24 shows the Factor Changes column property panel for the factor A1 in the Battery Data.jmp sample data table, located in the Design Experiment folder.
Figure A.24 Factor Changes Column Property Panel
• When a design contains factors that are hard-to-change and very-hard-to-change, it must also include a subplot and a whole plot factor:
‒ The levels of the whole plot factor define the groups of runs for which the levels of the very-hard-to-change factors are held constant.
‒ The levels of the subplot factor define the groups of runs for which the levels of the hard-to-change factors are held constant.
• When a design contains only factors that are hard-to-change, but no factors that are very-hard-to-change, it should include a whole plot factor. The levels of the whole plot factor define the groups of runs for which the levels of the hard-to-change factors are held constant. For more details, see “Changes and Random Blocks” in the “Custom Designs” chapter.
Augment and Evaluate Design
For the Evaluate Design and Augment Design platforms, the Factor Changes column property identifies factors with Changes specified as Hard or Very Hard. However, these platforms also require that the whole plot and subplot factors be entered as model effects in the launch windows. This is because the whole plot and subplot factors are part of the design structure.
Custom Design
The Custom Design platform enables you to create designs where all factor changes are Easy. You can also construct split-plot, split-split plot, or two-way split-plot (strip-plot) designs. When constructing these designs, you need to identify the factors whose values are hard-to-change or very-hard-to-change. In the Factors outline, you can identify factors as having Changes that are Easy, Hard, or Very Hard. When the Custom Design platform constructs the design table, the Factor Changes property is assigned to every factor that appears in the Factors outline.
The Custom Design platform is the only platform that constructs designs for factors with Changes that are Hard or Very Hard. Other DOE platforms also assign the Factor Changes column property to factors that they construct, but the value of the column property is set to Easy for their factors.
If you Load Factors in the Custom Design window using a table of factors, you can assign the Factor Changes column property to columns in that table. When you Load Factors using that table, your Factor Changes specifications appear in the Factors outline.
Factor Changes Example
The Battery Data.jmp sample data table, found in the Design Experiment folder, contains data from an experiment that studies the open current voltage of batteries (OCV). The design is a two-way split-plot design. For further background, see “Examples of Custom Designs” chapter.
1. Select Help > Sample Data Library and open Design Experiment/Battery Data.jmp.
2. Click the asterisk to the right of the factor C1 in the columns panel.
3. Select Factor Changes.
Figure A.25 Factor Changes Panel for C1
The value of Factor Changes for C1 is Hard. Figure A.24 shows that the value of Factor Changes for A1 is Very Hard.
4. Click OK.
5. In the data table, click the green triangle next to the DOE Dialog script.
6. Open the Factors outline.
Figure A.26 Factors Outline for Battery Experiment
The factors A1, A2, A3, and A4 have Changes set to Very Hard, and the factors C1 and C2 have Changes set to Hard. When the Custom Design platform constructs the design table, it saves these specifications to the appropriate columns as Factor Changes column properties.
In the Design outline, notice the Whole Plots and Subplots factors.
Figure A.27 Design Outline Partial View
To account for the factor changes that are Hard and Very Hard, two factors are created by the Custom Design platform. The Whole Plots factor groups the runs where the Very Hard factor levels are constant and the Subplots factor groups the runs where the Hard factors levels are constant. These factors need to be included as model effects when you enter columns with the Factor Changes column property in the Evaluate Design and Augment Design platforms.
Value Ordering
The Value Ordering column property assigns an order to the values in a column. That order is then used in plots and analyses. You can specify the order in which you want values to appear in reports.
Note: For certain values that have a natural ordering, such as days of the week, JMP automatically orders these in the appropriate way in reports. See The Column Info Window chapter in the Using JMP book.
Figure A.28 shows the Value Ordering panel for the Type column in the Car Physical Data.jmp sample data table. Reports that involve the values of Type place these levels in the order Sporty, Small, Compact, Medium, and Large. Use the buttons to the right of the Value Ordering list to specify your desired ordering for the values.
Figure A.28 Value Ordering Column Property for Type
In designs created using most DOE platforms, categorical factors, including the constructed factors Whole Plots and Subplots, and blocking factors are assigned the Value Ordering property. This property orders the levels according to the order in which they appear in the Factors outline. The levels of constructed factors are consecutive integers and the Value Ordering property specifies this natural ordering. You can modify the Value Ordering specification for any factor to meet your needs.
The Value Ordering property is not assigned by the Covering Array or Taguchi Arrays platforms. The Covering Array platform assigns the Value Labels column property. See “Value Labels”.
Value Ordering Example
Suppose that you want the values for a factor to appear in a different order in the Prediction Profiler. Consider an example of a wine tasting experiment, constructed using Custom Design. Wine is rated by five experts, each listed as a Rater in the Wine Data.jmp sample data table. Rater is a fixed blocking factor. Nine factors are studied. Rating is the response.
1. Select Help > Sample Data Library and open Design Experiment/Wine Data.jmp.
2. In the Table panel, click the green triangle next to the Reduced Model script.
3. Click Run.
The Prediction Profiler appears at the bottom of the report.
Figure A.29 Profiler with Original Value Ordering
Notice that the values for De-Stem and Filtering appear in the order No followed by Yes. You want to reverse these, so that the Yes level appears first.
4. Close the Response Rating report.
5. In the data table, click the asterisk next to De-Stem in the columns panel and select Value Ordering.
6. Click Reverse.
7. Click OK.
8. Click the asterisk next to Filtering in the columns panel and select Value Ordering.
9. Click Reverse.
10. Click OK.
11. Again, click the green triangle next to the Reduced Model script.
12. Click Run.
Figure A.30 Profiler with New Value Ordering
The levels for De-Stem and Filtering now appear in the order Yes followed by No.
13. Close the Wine Data.jmp sample data table without saving the changes.
Assigning Value Ordering
Consider the Candy Bars.jmp sample data table. Of the 18 brands lists under Brand, Hershey and M&M/Mars have the largest numbers of types of candy as listed in the Name column. You want these two brands to appear first in reports.
1. Select Help > Sample Data Library and open Candy Bars.jmp.
2. Select the Brand column.
3. Select Cols > Column Info.
4. Under Column Properties, select Value Ordering.
5. In the Value Ordering list, select Hershey.
6. Click Move Up five times.
Hershey is now first in the Value Ordering list.
7. In the Value Ordering list, select M&M/Mars.
8. Click Move Up seven times.
M&M/Mars is now second in the Value Ordering list.
9. Click OK.
An asterisk indicating the Value Ordering column property appears next to Brand in the columns panel. JMP now lists Hershey and M&M/Mars first in reports involving Brand.
10. Select Analyze > Distribution.
11. Select Calories and click Y, Columns.
12. Select Brand and click By.
13. Click OK.
14. While holding down the Ctrl key, from the red triangle next to Calories select Display Options > Horizontal Layout.
Note that the Distribution reports for Hershey and M&M/Mars appear first among the 18 brands.
15. Close the Candy Bars.jmp sample data table without saving the changes.
Value Labels
The Value Labels column property represents values in a column with specified labels. These labels are displayed in the data table and are used in plots and reports. In the data table, you can view the original values by double-clicking within a cell. For details about how to assign and work with the Value Labels column property, see The Column Info Window chapter in the Using JMP book.
The Covering Arrays platform is the only DOE platform that assigns the Value Labels column property. The Covering Arrays platform saves factors to the data table with a Nominal modeling type. The underlying values are consecutive integers ranging from 1 to the number of levels that you specify in the Covering Array Factors outline. The Values that you specify in the Factors outline are the Value Labels that are assigned to the underlying integers.
Value Labels Example
You want to test an internet-based software application to detect issues arising from components in the operating environment. The four components of interest consist of a browser, the operating system, the computer’s RAM, and the connection speed. To minimize testing time, you restrict yourself to testing two-way interactions.
Construct a Strength 2 covering array to test the required combinations of factor levels.
1. Select DOE > Special Purpose > Covering Array.
2. Select Load Factors from the red triangle menu.
3. Select Help > Sample Data Library and open Design Experiment/Software Factors.jmp.
The factors and their levels appear in the Factors outline.
Figure A.31 Factors Outline for Software Factors
Notice that the Role of the four factors is described as Categorical.
4. Click Continue.
The Restrict Factor Level Combinations outline opens. Since all combinations of settings are possible, leave this set to None.
5. Click Make Design.
6. Click Make Table.
7. In the columns panel, click the asterisk next to Web Browser and select Value Labels.
Figure A.32 Column Info Window for Factor A
Notice that Web Browser has a Numeric data type and a Nominal modeling type. The underlying values of Web Browser are the integers 1, 2, 3, 4, and 5. These values are assigned value labels corresponding to the values that you entered when you constructed the design.
Change the value label for 2 from IE to Internet Explorer.
8. Select 2 = IE in the Value Labels list.
Figure A.33 Value Labels Panel with Selection
9. Type Internet Explorer next to Label.
10. Click Change.
The change appears in the data table.
Note: To use the numeric values and not the labels, deselect Use Value Labels.
RunsPerBlock
When you use the DOE platforms to construct a design containing a blocking factor, the factor is assigned the Design Role column property with the value Blocking. JMP also assigns the RunsPerBlock column property to each Blocking factor. The RunsPerBlock property indicates the maximum allowable number of runs in each block. This property is used by the Evaluate Design and Augment Design platforms to indicate the blocking structure for the factor. For more details, see “Blocking” in the “Custom Designs” chapter.
Note: The RunsPerBlock column property is assigned by JMP as part of design construction. You cannot directly assign this column property.
RunsPerBlock Example
Consider the wine tasting experiment described in “Example of a Custom Design” in the “Custom Designs” chapter. Wine samples are tasted by five raters (Rater) and each rater tastes eight samples.
1. Select Help > Sample Data Library and open Design Experiment/Wine Data.jmp.
2. Click the asterisk next to Rater in the columns panel and select Design Role.
Notice that the Design Role is set to Blocking.
3. Click Cancel.
4. Click the asterisk next to Rater in the columns panel and select RunsPerBlock.
Figure A.34 RunsPerBlock Column Property Panel for Rater
Notice that the value of RunsPerBlock is 8. The design constructed by the DOE Dialog script has 40 runs. Since there are five raters, JMP constructs a design with 40/5 = 8 runs for each rater.
ConstraintState
In the Custom and Mixture Design platforms, you can Save Constraints that you specify for a design. When you select Save Constraints, the coefficients of each linear constraint appear in a column in a data table. The value that bounds the inequality is given in the last row of the table.
Each constraint column is assigned the ConstraintState column property. This property specifies the direction of the inequality that defines the constraint. When you select Load Constraints from a design platform, the ConstraintState column property tells JMP the direction of the inequality.
Note: The ConstraintState column property is assigned by JMP as part of design construction. You cannot directly assign this column property.
ConstraintState Example
The sample data table Piepel.jmp, located in the Design Experiment folder, contains a mixture design with three continuous factors. The design is based on an experiment presented in Snee (1979) and Piepel (1988), where there are boundary constraints on each factor and three additional linear constraints. In the following example, you do the following:
1. Change one of the three additional constraints
2. Save the constraints to a table
3. Observe how the ConstraintState column property describes the direction of the inequality in the constraint
For further development of this example, see “An Extreme Vertices Example with Linear Constraints” in the “Mixture Designs” chapter.
1. Select Help > Sample Data Library and open Design Experiment/Piepel.jmp.
2. In the Table panel, click the green triangle next to the DOE Dialog script.
Notice the three linear constraints below the Factors outline. To make the constraints more interpretable, you want to reformulate the first constraint in terms of a “greater than or equal to” inequality.
Figure A.35 Linear Constraints beneath Factors Outline
3. In the first constraint, do the following:
‒ Type 85 next to X1.
‒ Type 90 next to X2.
‒ Type 100 next to X3.
‒ Select ≥ from the inequality menu.
‒ Type 90 to the right of the inequality sign.
4. Select Save Constraints from the Mixture Design red triangle menu.
A table containing information about the constraints appears.
Figure A.36 Constraint Table
Each column contains the coefficients of the factors X1, X2, and X3 in rows 1 through 3. Row 4 contains the value that appeared to the right of the inequality sign.
5. Click the asterisk next to Constraint 1.
6. Click ConstraintState.
Figure A.37 ConstraintState Column Property Panel
The ConstraintState panel for X1 indicates that the direction of the inequality is “greater than” indicating greater than or equal to ≥.
7. Click Cancel.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.