************************************************************
The Unscrambler® X version 10.3
CAMO Software
Nedre Vollgate 8, N-0158, Oslo, NORWAY
May 2, 2013
************************************************************
What’s new in The Unscrambler® X version 10.3
================================
The following briefly describe the new features in version 10.3.
Design of Experiments
================================
- A completely new response surface plotting module with high resolution, fast graphics rendering and improved plotting controls for graphical optimization.
- A new D-optimal design module with option to augment design with space-filling points (more robust).
- Re-introduction of PLS-DoE and more design information displayed in ‘Tasks – Analyze – Analyze Design Matrix’ to help you find the best method for your data.
New methods
================================
- Basic ATR correction of absorbance transformed spectra included under ‘Tasks – Transform – Spectroscopic…’
- Introduced Double Kennard-Stone sample selection for PLSR, PCR and PCA (‘Mark’ menu from scores plot)
Plotting
================================
- Plot settings in ‘Tools – Options – Viewer’ can be used to change the default appearance of plots.
- New plots and plot layouts for Residuals and Influence plots in PCA, PCR, PLSR and Projection, including F-residuals with limits.
- Point labeling using value of any matching variable (Sample Grouping)
General
================================
- ASCII file import with default list separator based on system settings.
- New Alarms tab in analysis dialogs of PCA, MLR, PCR and PLSR and right-click option for setting alarm limits in the project navigator (these limits are applied for online prediction using some of our prediction engines).
- New dialog for assigning Scalar/Vector tags as well as units (‘Edit – Scalar and Vector’ in editor mode or right-click option in project navigator). This information is used for collecting data from various sources during online monitoring of processes.
- General enhancements and bugfixes.
************************************************************
The Unscrambler® X ver 10.2
CAMO Software
Nedre Vollgate 8, N-0158, Oslo, NORWAY
RELEASE NOTES
March, 2012
************************************************************
This document provides information about The Unscrambler® X ver 10.2.
The Unscrambler® X 10.2 contains several enhancements and new features for data import and export, analysis, graphics, and Design of Experiments (DoE). These
updates have been implemented post release of version 10.1. The Unscrambler® 10.2 is available in a 32-bit and a 64-bit version.
What’s new in The Unscrambler® X ver 10.2
================================
The following briefly describe the corrections and updates made in version 10.2
Applicability
========
Corrections have been made to address several issues:
- Overall performance of the program has been optimized, mainly based on the way data is stored in memory during calculations.
- More details of analysis methods and data have been added to info boxes.
- The **Find and Replace** functionality has been optimized.
- More time allowed for renaming project navigator nodes.
- The definition of Identity matrices in **Insert - Data Matrix** has been corrected to produce only square matrices.
- Median Absolute Deviation (MAD) scaling has now been moved to **Tasks - Transform - Centre and Scale** as a scaling option in the dropdown list.
- **Compute General** has been optimized to handle case-sensitive entries.
- **Audit Trails** now have a save option for printing and recording project details.
- In **Multiple Linear Regression** a rank dependency test has been added to better handle singularities.
- All analysis plots now have titles
Design of Experiments
===============
*New Features*
- Blocking of full factorial experiments
- Inclusion of up to 35 design variables in Plackett-Burman designs
- Design, response, and non-controllable variables are given in the same table for easier analysis
*Corrections*
- Overhaul of plots such as response surfaces, regression coefficients, multiple comparison, cube plots, residuals. Improved linking between plots.
- Optimized tests for consistency and simplex shape for mixture designs
- Better support for using actual values in mixture designs
- Model check section of ANOVA table will display sequential (Type I) sums of squares for mixture designs. Fixed linear terms DF.
- Rank testing implemented for quadratic, special cubic and full cubic terms in mixture designs.
- Optimized how non-constrained, mixture, and linear constrained variables are handled in DoE wizard
- Better handling of multilinear constraints for D-optimal designs
- Faster analysis of large Plackett-Burman designs
- Center samples inactivated if more than 4 levels for a categorical variable
*Known issues*
- The 9th design variable is by default called 'J', not the reserved letter 'I'
- Some larger fractional factorial designs removed due to large memory usage
- Upper limit imposed on the number of experimental runs for full factorial designs
- Display of B coefficients and effects plots/tables removed for designs with categorical variables with 3 levels or more.
- The DoE PLS option accessible from **Tasks -- Analyze -- Analyze design matrix** has been disabled. For D-optimal designs, use **Tasks -- Analyze -- Partial Least Squares Regression** instead. To analyze other designs using PLSR, change data type to numeric first.
- For models with category variables and centerpoints included, the total degrees of freedom is different from 10.1
Overall Enhancements
===============
*New Features*
- The **Import Interpolate** function has been included in some of the vendor-specific imports. This allows data to be combined into a single table if the starting and ending points are slightly different from each other.
- **Recalculate With New** is added as a new feature to PCA, PCR, PLSR, which allows the addition of new data to an existing model.
- Added new option for inserting Gaussian random data
- Improvements have been made in the way ASCII and Excel files are imported.
- Three new vendor specific imports have been included, DeltaNu, rap-ID and Visiotec.
- Export of files that can be used with the DeltaNu hand-held PharmID are now available.
- Digital signatures have been added to **File - Security** for enhanced data integrity purposes.
- A completely new implementation of the **Orthogonal Signal Correction** function has been made based on Tom Fearn's algorithm.
- An **Interpolation** function has been added to **Tasks - Transform**.
- A new algorithm for **Quantile Normalization** has been added to **Tasks - Transform**.
- **Mean, Min and Max** plots have been added to **Tasks - Descriptive Statistics** for better data visualization.
- Validation scores and leverages have been added to PCA, PCR and PLSR.
- Inclusion of validation residuals added to PCR, PLSR and MLR.
- **Support Vector Machine Regression (SVR)** has been added to the **Tasks - Analyze** menu.
- **Prediction** using SVR models is now provided in the **Tasks - Predict** menu.
- **Contingency Analysis** has been added to **Tasks - Analyze - Statistical Tests**.
- Plots have been added to the analysis node of **Linear Discriminant Analysis** and **Support Vector Machine Classification**
- **Compliance Mode** has been added as an installation option for those organizations that must meet the requirements of electronic signature and records handling.
- **Block Weighting** has been added as a new weighting option in the dialog boxes of PCA, MLR, PCR, PLSR, SVR, LDA, L-PLSR, SVC and is also available as an option in **Tasks - Transform - Weights**.
- **Q-Residuals** are now available at 6 levels of significance.
- A **Discard Residuals** option has been added to PCA, PCR and PLSR to help reduce the size of models based on large data sets.
- The **Search** option in **Help** has been optimized to include the following search features, *Match All*, *Match Any* and *Match Exact*.
- **Q-Residuals** are now available for analyses performed using the NIPALS algorithm.
- A function to show the Support Vectors is available for plots in SVC and SVR.
*User-Friendly Enhancements*
- The **Define Range** dialog has been completed overhauled for better ease of use and functionality.
- **Edit - Convert** allows the conversion of data collected in nanometers to be displayed in reciprocal centimeters (and vice versa).
- The **Fill** function is available as a right click option in the data editor.
- **Legend** and **Display Points** icons are now available in the toolbar.
- **Duplicate Matrix** is now available as a right click option in the project navigator.
- **Keep Outs** handling in all dialog boxes has been optimized.
- **Pretreatments** handling has been optimized in all analysis dialogs.
- An option for ordering variables in ascending or descending order in line plots has been added.
- Improved handling of batch importation of **ASCII** files
Known Limitations in The Unscrambler® X ver 10.2
===============
- When Uncertainty test is applied to PLSR, PCR, the uncertainty limits are provided for weighted coefficients only.
- Models with block weights are not compatible with v10.1. These models when used for recalculate in 10.1, will produce different weights.
Workaround: Reselect weights while recalculating
- If OSC was used as a transform in version 10.1, these values will not match those of version 10.2.
- p-Values of jack knife matrices will not match with v9.8. p-Value is set to 1 if the variables are down weighted
- Jack knife matrices will mismatch with v9.8 and v10.1.
- Correlation loadings will mismatch with v9.8 and v10.1 if weights are set to zero.
************************************************************
The Unscrambler® X ver 10.1
CAMO Software
Nedre Vollgate 8, N-0158, Oslo, NORWAY
RELEASE NOTES
January 12, 2011
************************************************************
This document provides information about The Unscrambler® X ver 10.1
The Unscrambler® 10.1 contains several enhancements and new features for data import and export, graphics, and Design of Experiments (DoE).These
updates have been implemented post release of version 10.0.1. The Unscrambler® 10.1 is available in a 32-bit and a 64-bit version.
What’s new in The Unscrambler® X ver10.1
===============================
The following briefly describe the updates made in version 10.1
Data Import
========
- The import of Excel data files did not always import all the columns from the Excel spreadsheet. This has been corrected.
- ASCII files can be batch imported
- U5 data can be imported into The Unscrambler®.
Data Export
========
File writer capabilities have been added to enable the export of models into The Unscrambler® 9.8 format. This feature can be downloaded from our website
www.camo.com
Applicability
========
Corrections have been made to address several issues:
- The axis labels in the influence plots in projection have been corrected to properly reflect the information plotted.
- Predictions made with models that include the MSC transform on part of the columns did not give consistent results. This issue has been addressed.
- Issues around the display of the correct sample names in the Coomans’ plot for classified samples using SIMCA have been resolved.
- The info box for a PLS model did not always correctly reflect the validation method used. When full cross validation was used, the validation was displayed
as having been random with 20 segments.
- The compute general function does allow mathematical formulae with non-integer values.
- Category variables can be copied and pasted.
- The x-axis values can now be scaled based on the variable values
- Compact, mini and micro models from previous versions of The Unscrambler® can be imported.
Design of Experiments
===============
Some changes have been made to how degrees of freedom are computed, relative to how this was done in previous versions.
Corrections have been made to address several issues:
- For experiments that include category variables, center points are defined for each level of the category variables.
- Response surface plots have been improved.
Overall Enhancements
- The grid editor has been modified to give improved performance. Data that are generated in version 10.1 cannot be opened in previous versions of The Unscrambler® X.
- Copy and paste and drag and drop have been implemented in the editor.
- In defining ranges, one can now define the reverse selections of the selected rows(columns) by a single click
- Prediction diagnostics per segment are available when a cross-validation other than full is used in developing PLS or PCR models
- The quantile normalization function of median absolute deviation (MAD) has been added as a new transform
- Q residual limits and Q residuals for samples are available within the results for predictions
- The ability to save model files as smaller files for easier model file transportability has been added.
- A user can set the number of components to save in a model file.
- Defined ranges (row and column) can be copied and pasted into a matrix of the same dimensions.
- The properties for category variables including the names and order of them can be changed.
- In LDA, the ability to do an automatic PCA-LDA on sample sets with many variables has been added under the options for LDA. Several improvements in the graphics
have been made including:
- Plot legends are now presented according to the sample grouping used in a given plot
- Users now have the ability to use sample grouping on 3-D scatter plots, and to more readily change the properties of 3-D plots
- Sample grouping is now possible under all the relevant PLS results plots including the X-Y relation outliers, Y-residuals vs. predicted Y and Y-residuals vs. scores
- With sample grouping, the groups in a plot can be separated by a symbol or a color, or both.
- Greater flexibility in changing plot and axis scales and labels.
- Additional options for the plot types have been added. Known Limitations in The Unscrambler® X ver 10.1
- Some issues still remain when calculating mixture designs with constraints, mainly due to summing of mixture amounts.
************************************************************
The Unscrambler® X ver 10.0.1
CAMO Software
Nedre Vollgate 8, N-0158, Oslo, NORWAY
RELEASE NOTES
July 19, 2010
************************************************************
This document provides information about The Unscrambler® X ver 10.0.1
The Unscrambler® 10.0.1 contains several enhancements for data import, graphics,
Design of Experiments (DoE) and security updates.These updates have been implemented
in response to customer feedback post first release of version 10.0.0.
What's new in The Unscrambler® X ver 10.0.1
================================
The following briefly describe the updates made in ver 10.0.1
Data Import
========
- GRAMS files with the `*.cfl` format can now be imported
- Import of BFF4 spectral files from Brimrose SNAP 32! Ver 3.01 is supported
- Import of `*.sp` and `*.spp` files from Perkin Elmer Spectrum software is supported
Tutorials
======
Additional tutorials have been added to give users a quick start in using
The Unscrambler® X
The ability to mark evenly distributed samples in scores plots and other PLS results
plots has been added
Applicability
========
Several improvements in the graphics have been made including:
- Sample grouping can be applied in 3-D plots
- When a sample or variable has been marked in a plot, it can now be unmarked
by clicking again
- Sample grouping can be applied to DOE data plots, influence plots, and other
plots as relevant
- Plot names have been updated to Predicted vs. Reference (previously Predicted vs. Measured).
- Data can be specified as spectra, thus showing loadings and regression coefficients results as line plots
when relevant, by selecting a column set, right clicking and specifying spectra
- The plots for PLS and PCA analysis are displayed with the optimal number
of factors for the given model
- Plots for PLS prediction are shown with the number of factors used in the model
- Response surface plots have been updated
- The Y axis in a plot can be changed by going to Properties
Corrections have been made to address several issues:
- Import of ASCII file format has been corrected
- When models are developed on data that have an MSC or EMSC transformation
applied, the transformation is registered and applied for predictions
- The log transformation in Compute General works as expected
- Issues related to incorrect variable names on import of OPUS files has been
corrected
Design of Experiments
===============
Updates have been made to the Design of Experiments, with results confirmed with
reference to D. C. Montgomery "Design and Analysis of Experiments" 6th ed.
- ANOVA can be run when category variables are present in a design
- Corrections have been made to the experiment names when a fractional
factorial design is generated
- Updates have been made to the Mixture design, and it has been modified so it
can be done with 3 or more variables only
- The x-axis labels in the B coefficient plots correctly reflect the variable names
Known Limitations in The Unscrambler® X ver 10.0.1
=================================
- The OSC transformation has been modified, hence the old OSC model (from
version 10.0) cannot be used as a registered pretreatment in prediction.
************************************************************
The Unscrambler® X ver 10.0.0
RELEASE NOTES
April 10, 2010
************************************************************
What's new in The Unscrambler® X
=========================
Savitzky Golay Smoothing - This function has been slightly changed
from previous versions of The Unscrambler®. The end points of the samples
now keep the original data instead of filling with zeros.
The Median Filter algorithm in The Unscrambler® v9.8 produced incorrect results.
This has now been fixed, by taking the median value of point pairs less than the size
of the window used.
When applying a Gaussian Filter, the columns are shifted by 1 so that the peaks are
not repositioned, as they were in The Unscrambler® v9.8.
SIMCA in The Unscrambler®X is not calculated for full rank.
The Varimax Rotation for PCA produced incorrect results in v9.8. This has
been corrected in The Unscrambler® X.
Kurtosis and Skewness values computed in The Unscrambler® v9.8 and
earlier versions do not match with MS Excel or Matlab.
The Unscrambler® now implements skewness and kurtosis with bias
correction which is used in
Microsoft Excel `skew()` and `kurt()` functions;
Matlab functions `skewness(x,0)` and `kurtosis(x,0)`.
In the first release of The Unscrambler® X, the import of 3-D data is not supported
and neither is N-PLS.
Sample grouping in 3D scatter plots is not available in the initial release.
When a sample or variable is marked in a plot, it cannot be unmarked by clicking again.
Some of the HCA distance measures like Spearman's are not identical with a test
implementation validated in using an external source.
Annotations and sample groupings of plots are not saved in a project.
The recalculate option is not available in descriptive statistics.
NIPALS incompatibility - In The Unscrambler® 9.8 and previous versions,
the NIPALS algorithm has been implemented using two techniques.
The NIPALS for calibration is different from the one used during prediction
and classification. The Unscrambler® X uses single NIPALS algorithm for
calibration, prediction and classification.
In Unscrambler® v9.8 and previous, the square sum of scores includes test set
samples. This is wrong. The square sum of scores should include only the calibration
set and not the test set. This has been corrected in The Unscrambler® X
The Analysis of mixture designs is now performed by the use of Scheffe polynomials
instead of PLS, as in previous versions. The PLS option is still available.
For the initial release of The Unscrambler® X, for category variables, only the
multiple comparison results are provided as in 9.8. The real effect values will be
provided in the next release.
The handling of constraints has been improved in D-optimal designs. A wider
range of operators is available.
In The Unscrambler® 9.8 and previous versions, the definition of the design,
response, uncontrollable variables was made in three different windows.
This has been reduced to the Define Variables table in the Design Experiment Wizard.
Method reference documentation is yet to be updated.
System Requirements
==============
for Standalone and Client
* Windows 2003, XP (SP1 or higher), Vista, 2008, 7 or later
* Processor: Pentium 233-megahertz (MHz) or higher
* RAM: 1 GB minimum; higher is recommended
* Hard disk space: 150 MB
* .NET framework versions 2.0 to 3.5
* Internet Explorer 7.0 or higher
for License Server
* Windows 2003, XP (SP1 or higher), Vista, 2008, 7 or later
* Processor: Pentium 233-megahertz (MHz) or higher
* RAM: 1 GB minimum; higher is recommended
* Hard disk space: 50 MB
* Internet Explorer 7.0 or higher
Installation
=======
1) Run The Unscrambler® X setup application & follow the setup wizard
Double click "TheUnscramblerX_Setup.msi" file to start the installation wizard.
The InstallShield Wizard for The Unscrambler® X is launched. Follow the
on-screen instructions
2)Finish the Setup. When the setup is complete, Click
3) Start The Unscrambler® X from the Start menu
4) Step 4 : The Activation Wizard dialog opens.
Click the button
5) After receiving The Unscrambler® X activation key, paste in Activate
window and click on button
OR
Send “machine ID” from The Unscrambler® X Activation window along with
your user name and E-mail address to support@camo.com.
CAMO Support Team will send you The Unscrambler® X activation key.
************************************************************
For Support visit : http://support.camo.com
************************************************************
CAMO Software
Research & Development Team