Design-Expert^® version 8
software for experiment design

July 14 , 2011

ELECTRONIC SOFTWARE DELIVERY (ESD)

If you downloaded your Design-Expert 8 purchase via ESD, please read these installation notes before installing it: www.statease.com/files/InstallESD.htm.

ON-LINE REGISTRATION

If you purchased the full version of Design-Expert (single user or network license) and haven't registered, you may do so on-line 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 Design-Expert 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 Design-Expert program, choose Help/Contents from the menu bar. Expand the topic node, "Welcome to Design-Expert 8," and you will find a sub-topic 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:

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)

Design-Expert 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, Design-Expert 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 Design-Expert 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.

July 14 , 2011 - released version 8.0.6

Release version built and tested.

Corrections and changes made in version 8.0.6

Visual C++ 2005 SP1 Redistributable Package: 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 one-sided In-Range 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 user-definable 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 1-99.
• 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 user-added 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 tab-delimited 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 Half-Normal plot we now copy the effects to the clipboard when control or shift right-clicking on Half-Normal 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 y-axis settings
• Auto-adjust 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 right-clicking 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.
• Box-Cox handles multiple inflection points better.
• Points were colored incorrectly in Graph Columns for U-Pseudo mixture components when colored by factor.  They are now colored correctly.
• Refined the plotting of effects not in the model for Multilevel Half-Normal 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 non-mixture 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 non-orthogonal factorial designs.
• Half-Normal plot no longer drops a term when some are set aside.
• Y-axis 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 right-clicking on the column header. For mixture components, the coding can be switched between U-Pseudo and L-Pseudo.
• 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 split-plot 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 ease-of-use 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 2^6-1 design with 4 blocks.
  
• Added a progress bar to 10 component special cubic simplex lattice build with 5 blocks.
  
• Eliminated unnecessary U-Pseudo 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 Mix-Mix 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 one-sided confidence, prediction or tolerance intervals in a different color.
• User-entered 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 all-discrete 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.

U-Pseudo coding

• U-Pseudo axes values are now properly labeled on all graphs.
• The user can now change U-Pseudo coding to L-Pseudo when rebuilding a U-Pseudo mixture.
• Various improvements to handling U-Pseudo 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 half-normal 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 Design-Expert 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 User-Defined 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.
• Half-normal 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 non-hierarchical 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 double-clicking 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 non-factorial 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:

www.statease.com

* replace _at_ with @ to use the e-mail address.

Design-Expert^® is a registered trademark of

Stat-Ease, Inc.
2021 East Hennepin Avenue
Suite 480
Minneapolis, MN 55413

Copyright © 2010. All rights reserved.

ACKNOWLEDGEMENTS