DesignExpert^{®} version 8
software for experiment design
December 30, 2011
ELECTRONIC
SOFTWARE DELIVERY (ESD)
If
you downloaded your DesignExpert 8 purchase via ESD, please read these installation
notes before installing it: www.statease.com/files/InstallESD.htm.
ONLINE
REGISTRATION
If
you purchased the full version of DesignExpert (single user or network license)
and haven't registered, you may do so online at www.statease.com/dx_register.html.
WHAT
IS NEW IN VERSION 8?
A
description of what's new can be found in the help system. In the program, choose
Help/Contents from the menu bar. Under the topic, "Welcome to DesignExpert
8" you will find a section titled, "What's new in version 8 (the
highlights)" with a section for version 8.
FREQUENTLY
ASKED QUESTIONS
These can be found in the Help system. From the DesignExpert program, choose
Help/Contents from the menu bar. Expand the topic node, "Welcome to DesignExpert 8," and you will find a subtopic titled "Program Hints and Frequently
Asked Questions."
INSTALLATION
NOTES
Supported
operating systems
Windows XP
Windows Vista
Windows 7
System
Requirements for single user and network clients:
Attribute 
Minimum 
Recommended 
Display
resolution 
800x600 
1024x768
or greater 
Memory
(XP/Vista or 7) 
256/512
MB 
1
GB/2 GB or greater 
Processor 
Pentium
IV 
Pentium Duo Core or greater 
Operating
system 
Windows
XP 
Windows
XP, Vista, Windows 7, or later 
Disk space (program/free) 
40 MB / 5GB 
Copious free space for virtual memory 
Server requirements for LAN networks:
The version 8 server component requires a file
server with at least 50 MB available disk space. There are no
components or services that run on the server, but users must have
read / write access to a control folder that you designate.
Math
Kernel Library (MKL)
DesignExpert 8 uses Intel's Math Kernel Library (MKL) to optimize the computational
speed. It is distributed as part of the installation. Starting with version 8.0.3, DesignExpert uses version 10.2* of the MKL. Earlier releases used version 7.2 of the library. Should your system be incompatible with this library, it can be disabled in the Math preferences tab from the main menu.
Help
file on network
Windows XP workstations:
A security change in XP prevents it from opening HTML help files that reside
on a LAN or the Internet. If the dx8HelpFile.chm file doesn't work from your
network drive see the Microsoft support document, http://support.microsoft.com/?kbid=896054
To permit LAN access follow the instructions in
Example 2 and set the MaxAllowedZone value to 1.
If you use WrkSetup.exe to set up DesignExpert on you workstation, it will set the MaxAllowedZone for you.
Windows Vista/Windows 7 workstations:
Windows Vista and Windows 7 do not allow HTML help to be used over the network. To have the help file available you must install the help file on your workstation. WrkSetup.exe provides an option to install help locally.
KNOWN ISSUES
If
you are not able to activate your software via the internet, you may need to temporarily disable your firewall.
December 30, 2011  released version 8.0.7.1
Release version built and tested.
Corrections and changes made in version 8.0.7.1
 Corrected problem with estimating power for interactions with categoric factors.
 XML script command SetResponseModel now deletes any previous analysis, forcing a reanalysis with the new model.
 Expanded version number displayed throughout program to include all four identifying numbers.
December 12, 2011  released version 8.0.7.0
Release version built and tested.
Corrections and changes made in version 8.0.7.0
 Added a notation for graphs and prediction reports of transformed responses in Original Scale that the estimates are for the median not the mean.
 XML now exports and imports Equation Only responses and interval data used in optimization.
 Added terms can now be entered when building all types of combined designs and will be restored to the design model when rebuilding.
 Corrected some problems graphing and optimizing mixtures with invariant components.
 Eliminated incorrect error message which sometimes prevented displaying Graphical Optimization Graphs for four component mixtures.
 Improved robustness rebuilding optimal designs with discrete numeric factors.
 Improved accuracy when calculating: ShapiroWilk in Effects HalfNormal plot; Standardized Effects for error terms not in the Design model; and POE for logtransformed responses.
 Minor tweaks to the interface and memory management.
July 19 , 2011  released version 8.0.6
Release version built and tested.
Corrections and changes made in version 8.0.6
Multilevel Effects Selection
 Improved Bonferroni preselection to avoid sometimes overselecting terms.
 On the HalfNormal plot, changed the adjustment of normal effects that are too large to be estimated with standard Chisquared function. Now we scale to largest estimable effect based on the mean squares.
 ShapiroWilk values are no longer displayed in multilevel HalfNormal plot.
Intervals
 Fixed a problem with interval calculations for transformed response in original scale when using the POE adjustment.
 Now report POE intervals in Point Prediction when valid but standard deviation set to zero.
 Improvements to tolerance interval calculations for extreme cases such as when extrapolating outside the design space.
Graphs
 Eliminated some anomalies switching axes in Overlay graphs with categoric factors and 3D Model graphs for mean only models.
 Corrected indexing problem when highlighting points with endpoints are not displayed in XY Graphs.
 LSD bars were incorrect when the coding for analysis was set to Actual. Fixed.
 Corrected the legend for intervals which switched the line style for prediction and tolerance intervals.
 More precise calculation of intersection of FDS and Std Error lines with FDS curve on FDS graph.
 Flags planted on Evaluation contour graphs had some missing values, now these are displayed properly.
Miscellaneous
 Eliminated program crash in Annotated ANOVA when analyzing aliased model in Actuals.
 Adding a factor to combined design no longer results in incorrect evaluation of constraints.
 Reading in a candidate set when building an optimal mixture could result in an incorrect mixture total. Resolved.
 Made a number of minor tweaks such as centering message boxes in the application window and improved text.
 Updated Visual C++ 2005 SP1 Redistributable Package to the June 14, 2011 Security update.
March 31, 2011  released version 8.0.5
Release version built and tested.
Corrections and changes made in version 8.0.5
New Features
 Optimization
 Added Confidence, Prediction and Tolerance Intervals to Numerical Optimization for Maximize, Minimize and In Range response goals. The desirability function will now take the intervals into account. Also, response models now allow for onesided InRange goals: if one of the limits is missing, the response will be unconstrained on that side.
 More robust checking for unvarying range in Std Error models in Numerical Optimization. Before we only checked design points, now we add the centroid if not present and high and low are equal.
 Graphical Optimization graphs now display active responses and intervals in the legend. Also, allow user to delete response contours which removes the corresponding limit in criteria. The interval contour label text has been shortened, too, and the interval options no longer shows up for equation only responses. An userdefinable Outside Limits color is now included (see below).
 XML scripting
 XML scripting now supports an XMLDiff command which will compare two XML files and a ScriptFailureAlert command which will display a success or failure message after a script has been run. This allows the program to run validation tests in XML by comparing XML report outputs with answer files. Such a test (SelfTest.xml and accompanying files) is now provided with the program under the SelfTest directory.
 Added export confirmation report command.
 Preferences
 User can now choose font sizes from 199.
 A new color preference (Outside Limits) added for areas outside the limits but inside the constraints for Graphical Optimization overlay graphs (see above).
 A Browse button is available for the Pathnames options in the Data Files tab of Advanced Preferences. Default document pathname is now set by entering the Data directory and the document name separately.
 Advanced Preferences contains a new Math tab with five options including three new ones:
 Include Linear+Squared terms and Enable Partial Quadratic Mixture (PQM) models in mixture designs of 4 or more components. This feature was always enabled previously and now is disabled by default.
 Set the exported precision of numbers in XML and CSV (comma separated value) format reports (new feature).
 Display POE adjustment for intervals when available (new feature, see below).
 The existing Use MKL routines option and Show orthogonal direction on mixture ANOVA and Trace plots option have been moved from the Math tab of Preferences.
 Replicates
 We have improved the replication of runs in blocked mixture simplex designs. Before, when building mixture simplex designs, replicates were placed in separate blocks which sometimes resulted in zero degrees of freedom for pure error estimation. Now we split the replicates into two groups, those needed for block estimation (added prior to blocking) and those needed for pure error (added after blocking). The first number is calculated by the program and reported on the build screen, the second number is entered by the user.
 Better replicates in Optimal designs using Modified Distance criteria. Previously, we scaled the Modified Distance criteria by D to avoid singularity but sometimes the D would dominate the distance criteria. Now we attempt to find a local Pareto optimum, i.e., neither the D nor the Modified Distance criteria are allowed to decrease.
 Propagation of Error
 If POE is calculated, we now adjust the intervals displayed in graphs for variation transmitted by the factors.
 POE graphs were disabled in Model Graphs when POE model was constant. Now they are enabled.
 Improved the accuracy of POE calculation for cubic and higher order Scheffe terms for mixtures and combined designs.
User Interface
 Corrected a couple of problems where scrollbars disappeared.
 Improved model selection by deselecting useradded terms when model is selected by order.
 Eliminated duplicate block contrasts warnings when opening a design.
 The Edit Constraints dialog sometimes would not parse a coefficient in exponential format. This has been fixed.
 Corrected a problem that arose in version 8.0.4 with copying and pasting tabdelimited text.
 Switching to sheet view in Factors tool sometimes could change component settings. This has been fixed.
 Missing factor levels are now preserved with when changing categoric levels in Edit factor dialog.
 In Effects HalfNormal plot we now copy the effects to the clipboard when control or shift rightclicking on HalfNormal plot in addition to displaying them in a menu for selection.
 Fixed problem with updating alpha edit fields in CCD build when entering factor ranges in terms of alpha.
Graphs
 Improved default response yaxis settings
 Autoadjust option is now off by default, when off, the axis range now only expands, never contracts.
 Response axis format now defaults to current response format not General and is editable in more situations.
 More consistent handling of axis settings across different types of graphs including transformed responses in Original scale and Numerical Optimization response graphs.
 Better default selection of axes for Main Effects model with numeric and categoric factors and when switching between Contour, 3D, and One Factor graphs for Combined designs.
 Fixed problem with reference line being misdrawn on graphs of a categoric factor when dragged to the edge.
 Corrected some minor problems with messages related to design points on 3D graphs. Also, fixed a crash that could occur when rightclicking outside constraints in mixture contour graph when design points are turned off.
 Fixed problems with displaying the requested number of grid lines in One Factor and Interaction graphs and with labeling grid lines in Actual rather than Pseudo coding in Mixture Contour graphs.
 BoxCox handles multiple inflection points better.
 Points were colored incorrectly in Graph Columns for UPseudo mixture components when colored by factor. They are now colored correctly.
 Refined the plotting of effects not in the model for Multilevel HalfNormal plots.
Reports
 New improved text for Annotated ANOVA when curvature is selected in Effects Model.
 Changed the labeling of Low/High Actual/Coded values in Design Summary for clarity.
 More efficient and robust calculation of combined models in Fit Summary.
 The ANOVA was reporting very small sums of squares as zero. Now it reports the actual values.
Miscellaneous
 Added Point Types to One Factor RSM designs.
 Eliminated a program crash in Evaluation when trying to evaluate a nonmixture model with no intercept term.
 Corrected indexing problem with augmenting a design with ignored runs. Also, fixed a crash augmenting or rebuilding an augmented design with blocks but no replicates using IV criteria.
 Remedied error where number of replicates was sometimes incorrect when rebuilding factorial designs.
 Reopening a saved and analyzed mixture design with invariant components could report bad component ranges in Numerical Optimization. This has been corrected.
 Eliminated a couple of crashes in Effects with aliased designs (including Forward selection of a multilevel categoric factorial model with terms that are only partially estimable).
September 15, 2010  released version 8.0.4
Release version built and tested.
Corrections and changes made in version 8.0.4
New Features
 A Confirmation Report is now available for calculating intervals around confirmation points.
 The Color by option is now listed in the legend for Graph Columns and has improved indexing with missing data.
 Runs added by augmenting are now assigned a random run order (the run order for the previous runs is unaffected).
 Can now read in a candidate list with discrete numeric factors in Optimal builds. Also, handles reading in all categoric candidate points properly.
Building a design
 Constrained RSM design with discrete factors no longer builds with points outside the constraints.
 Eliminated a crash that could happen when you create a General Factorial design.
 Better bootstrap generation for Combined designs.
Graphs
 Eliminated crashes that occurred in some graphs for Mixtures and Combined designs.
 Improved handling of transformed responses in graphs.
 Clearer wording in warning message for nonorthogonal factorial designs.
 HalfNormal plot no longer drops a term when some are set aside.
 Yaxis label text was occasionally getting truncated. Fixed.
 Graph preference settings for format and ticks in Mixture designs are now applied more consistently.
Miscellaneous
 The FPDS graph in Evaluation now gives more accurate results for all categoric designs in Evaluation.
 Resolved problem calculating categoric actual equations for analyses with POE model and more than 3 categoric factors.
 Better copy/paste feature in Coefficients table.
 Categoric level was sometimes mislabeled in Optimization Ramps. Fixed.
Corrections and changes made in version 8.0.3
New features
 New Recode factor option on the design layout allows user to change the coding for a factor based upon the data entered by rightclicking on the column header. For mixture components, the coding can be switched between UPseudo and LPseudo.
 Advanced preferences dialog is now available under the Edit menu. This allows the user to save, load, reset, and refresh preferences and to edit pathname and Vartran choices.
 Improved handling of Block and Set Aside models for nonorthogonal splitplot and nested designs.
 Prediction graphs now display transformed response in original scale by default.
 Point Prediction Report now displays the standard deviation used to calculate the other columns for a response with a notation if it was entered by the user.
 Updated Intel Math Kernel Library to version 10.2 and reduced the library included in distribution to only those functions used by the program. This reduced the size of the installation program by about 10MB.
User Interface
 Improved the easeofuse of the Insert factor or response function in the design layout.
 When editing the factor low and high on the design layout we now skip the adjust actual values warning message if the actual low is 1 and high is +1 and automatically adjust the actual values.
 Fixed some minor problems with entering standard deviation for a factor or response.
Building a design
 We now add blocks before replicates in Optimal builds, which produces a better design.
 Fixed a number of problems involving rebuilding optimal designs.
 We now warn user when adding center points to an all discrete RSM Optimal designs.
 Improved performance for User Defined and Optimal combined designs.
 The fractional factorial grid now displays the correct resolution color for 2^61 design with 4 blocks.
 Added a progress bar to 10 component special cubic simplex lattice build with 5 blocks.
 Eliminated unnecessary UPseudo warning when switching ranges back to L_Pseudo.
 Initial default candidate point checkboxes for User Defined design are now always updated when the model changes.
Analysis
 The program no longer hangs doing backward selection on aliased quadratic mixture model.
 We now skip reporting Adjusted model Lack of Fit in ANOVA Summary table if it has no degrees of freedom. Also, we skip the Summary table entirely when there is nothing to report.
 Eliminated a crash analyzing General Factorial with missing data.
Graphs
 Improved export of graph labels.
 Export to PNG or JPG format did not draw background behind legend, now it does.
 Improved display of design points in all graphs.
 Continued improvements to graphs for transformed responses, for example, we now use original rather than transformed scale for color on contour and 3d graph.
 Fixed some minor problems with crosshairs windows values.
 Added a graph preference item for exporting gradient fill shading. The default is off.
 Confidence bands and LSD bars are now more accurate for One Factor categoric and Two Component Mix graphs.
 More consistent formatting of factor and response values.
 Can now switching graphs using 3 factor terms from term list for combined mixture by process designs.
 A number of problems fixed and improvements made to the display of X and Y axes.
 Eliminated several minor glitches with solution and planted flags in Model and Optimization graphs.
 Improved perturbation and trace plot displays which is now also available for POE model.
 Surface Graphs tab added to Graph Preferences for Contour MixMix graph.
 Removed Pred v Actual choice for Equation Only Model graphs.
 Graphical Optimization overlay mix graph coloring of areas outside constraints was made to be consistent with other graphs.
 Added d and s values to legend on FDS graph.
 Numerous minor fixes to improve appearance and consistency of graphs.
Optimization
 Better formatting of response interval values in Point Prediction.
 Corrected some problems exporting the optimization in XML.
 Numerical Optimization Prediction value is now uniform in all graphs.
 All factors can now be set to be "equal to" in Numerical Optimization.
Miscellaneous
 Better rounding of confidence, prediction and tolerance interval percent values.
 The Augment Design feature is more robust and has had a number of minor improvements made to the interface.
 Added a warning to legend that not all terms are available when viewing DFBETAs with multilevel terms present in the model.
 Fixed a crash that sometimes occurred when canceling on Save file.
April 12 , 2010 version 8.0.2
Release version built and tested.
Corrections and changes made in version 8.0.2
New features
 Added Interval regions for responses in Graphical Optimization. Now the overlay graph will display areas that do not meet onesided confidence, prediction or tolerance intervals in a different color.
 Userentered standard deviations are now used throughout the program in place of the ANOVA calculated value, e.g., the std error graphs and the confidence, prediction, and tolerance intervals in the graphs, Optimization and Point Prediction.
 Point Prediction now displays POE interval data when available.
 The ability to read in a candidate set was not working, it has been restored.
Increased robustness
 Removed potential collisions with Ultramon.
 Eliminated a number of crashes that occurred when analyzing or augmenting unusual factorial designs.
 Program now handles incomplete designs with missing runs and those with a single level better.
 The Coefficients Table now works for mixture designs with blocks.
 Improved processing of messages from Factors Tool in contour graphs for large designs.
 Design layout no longer gets stuck in recurrent warning dialogs when adjusting runs in a mixture with invariant components.
 Fixed crash that happened occasionally when selecting New Design on welcome screen.
 Problems augmenting designs with categoric factors have been addressed.
 Corrected a number of problems with aliasing and model reduction for fractional factorial and multilevel categoric factors.
Building a design
 Improved coordinate exchange selection of designs with categoric factors.
 Fixed various problems with building or rebuilding combined designs .
 Eliminated crash when rebuilding modified CCD with blocks.
 Sped up General Factorial build with 2 replicates and no blocks.
 Enhanced point selection for linear RSM optimal design.
 Corrected problem with alldiscrete RSM optimal designs where points selected were out of range.
 Fixed some scattered user interface glitches.
Transforms
 Row label now displayed properly for transformed response in Coefficients Table.
 Legend in Diagnostics is now always in transformed scale.
 Negative power transforms no longer give incorrect desirability values in Numerical Optimization when in original scale.
UPseudo coding
 UPseudo axes values are now properly labeled on all graphs.
 The user can now change UPseudo coding to LPseudo when rebuilding a UPseudo mixture.
 Various improvements to handling UPseudo coding throughout the program.
Graphs
 Relabeled some postscript driver files so the program can export graphs in EPS format.
 Reduced unnecessary recalculation of graphs.
 Got rid of unneeded warning messages and refined the wording on others.
 Now have better defaults for axes for Graph Columns, Diagnostics, Model Graphs, and transformed responses.
 Improvements to Factors Tool for both gauges and sheet view.
 Minor adjustments to legend for different graphs.
 Better reporting in crosshairs window and planted flags.
 Rectified other small defects in graphs.
 Removed scale error option from halfnormal graph in Effects.
User Interface
 Design layout does not return to first column after resizing.
 Selecting Optimal augmentation no longer causes DX8 to lose focus briefly.
 Corrected a number of minor problems with scrollbars in factor tables and Optimization reports.
 Inserting a new factor no longer gives a warning about adjusting coded/actual levels.
 There is now better sizing of progress bars.
 There have been been several changes to clean up minor problems with the reports.
Miscellaneous
 Changed the default tolerance population proportion value to 0.99.
 Add blocks drop down now prevents user from adding more blocks than rows.
 Updated XML interface.
Installation (All)
 Added option to create desktop icon for DesignExpert 8.
 Optional automatic checking for program updates is now available.
 Removed gdiplus.dll from the installation. It's functionality is included in Windows XP and later versions.
Installation (Single user only)
 Single user activation is no longer available during the install.
Installation (Network client only)
 New features added to the automated workstation setup (See the document, http://statease.com/files/Network/WrkSetup802.pdf, for details.)
 Install desktop icon for all users or current user only.
 Set/disable automatic update alert and frequency.
 Totally silent install of Microsoft C++ 2005 redistributables. (A tip of the Trigger Topper to Jerome McCann for this implementation.)
 Corrected a number of bugs with previous automated install.
January 20, 2010 version 8.0.1
Release version built and tested.
Corrections and changes made in version 8.0.1
Building a design
 Eliminated a number of crashes that occurred when switching between various designs and study types.
 Missing scrollbars for the factors entry table now appear when needed.
 Categoric factors will now no longer appear in Constraint Tool dialog as they cannot be part of a constraint.
 Factor information that was sometimes lost when rebuilding designs is now preserved.
 The number of additional center points added when categoric factors are present is now reported correctly.
 Can now build a mixture optimal design with an invariant component.
 Default checkboxes for candidate points in UserDefined designs are now being set according to the selected model.
 Forcing categoric balance in an optimal design did not always result in balanced designs. That has been corrected.
Design Layout
 Can now delete a numeric factor from design layout in a combined design.
 Improved factor/component adjustments when runs do not meet constraints.
 Program will now prompt user to save when closing design after a comment is added or changed.
Evaluation
 New annotation for VIFs in Evaluation Report.
 Evaluation model was sometimes wrong when a response was switched from Polynomial to Factorial analysis. Corrected.
Effects
 Fixed a number of crashes in Effects model selection for unusual designs.
 Halfnormal plot was incorrect after selecting a model when analyzing in Actuals. Fixed.
 For a one factor categoric design with pure error user can now select model by probability.
ANOVA
 Crash when turning off actual equation in nonhierarchical ANOVA when analyzing in actuals has been corrected.
 Actual equation coefficients were using adjusted ANOVA instead of unadjusted. Fixed.
 The change Sum of Squares warning has been reworded to be less confusing.
Graphs
 Program no longer exits under XP when accessing help in Diagnostics Pred v Actual.
 Surface graph axes labels are now updated properly when changing the component settings for mixtures with more than 3 components.
 Fixed some minor problems with flags on 3D graphs.
 To zoom in the user must now hold down the control key when doubleclicking the left mouse button. To zoom out, the user must also hold down the shift key.
 Diagnostic graphs now update ticks and axis range when switching between Studentized and Unstudentized.
 Improved default range for Standard Error graphs when response is transformed.
 Reference line SnapToGrid preference now defaults to off.
Optimization

Improved numerical optimization when setting criteria for a factor to a single categoric level.
 Improved handling of transformed responses in Numerical and Graphical Optimization.
 Numerical optimization mix component goal values are now remembered for equal to criteria.
XML
 A number of minor problems with the XML output and scripts were corrected.
 The XML design document is now always saved in run order. Old files will read in properly in the new version but standard order will be altered when reading new files in with the old version.
Miscellaneous

Factorial augment no longer gives nonfactorial points with coordinate exchange.
 Entries in the Coefficients Table were not always being colored correctly or displayed in proper order. Fixed.
 Point Prediction report did not always resize when response analyses were changed. Corrected.
INTEL MKL 10.2
TECHNICAL
SUPPORT
For
Technical Support or other information contact:
*
replace _at_ with @ to use the email address.
DesignExpert^{®}
is a registered trademark of
StatEase,
Inc.
2021 East Hennepin Avenue
Suite 480
Minneapolis, MN 55413
Copyright
© 2010. All rights reserved.
ACKNOWLEDGEMENTS
1. 
DesignExpert 8 uses computational routines from LAPACK (Linear Algebra PACKage), Copyright © 19922008 The University of Tennessee. All rights reserved. 

LAPACK carries the following notice: 

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 