The Unscrambler® X with Design-Expert®
CAMO software has teamed up with Stat-Ease to provide customers with access to their leading Design-Expert® software for experimental design. The Unscrambler® X version 10.4 or 10.4.1 with Design-Expert provides users with access many powerful statistical tools and experiment designs to optimize their products or processes. Supported versions are version 9 and the newly released version 10.
DOE features include:
- Two-level, general and optimal factorial split-plot designs.
- Enter verification runs embedded within blocks as controls or appended to your completed design
- Adjustably-tuned LOESS fit line for Graph Columns.
- Enter a single factor constraint for response surface designs
- Power calculated for split plots versus the alternative of complete randomization
For a full list of features, click here.
The first section below briefly describes other new features in The Unscrambler® X v10.4.1, while the following sections describe features included since v10.4.
Changes in v10.4.1
- Moving Block Methods with F-test limits and plots.
- Contribution plots added to PCA, PCR and PLSR models.
- General enhancements and bug fixes.
New methods in v10.4
- ‘Statistical Process Control’ (SPC) has been implemented for univariate monitoring of e.g. manufacturing processes.
- ‘Moving Block Methods’ provides simple SPC charts for univariate or multivariate trending of process data.
- ‘Piecewise Direct Standardization’ has been included to enable calibration transfer.
- ‘Sample Alignment’ has been developed to merge two or more data tables based on sample time stamps or category levels.
- The ‘Reduce (Average)’ transformation has been modified to allow group averages based on sample indexes, and to handle non-numeric data.
- Bias and Slope Correction is available for regression models.
- A ‘Split Text’ function has been added that allows splitting of sample names or text columns based on a text separator or character positions. This is available from the Edit menu.
- Option to augment Kennard-Stone sample selection to achieve more uniform distributions of response values has been added.
- Modified ‘Matrix Calculator’ dialog, including a new Category Reshape function.
- Added ‘Design-Expert’ import for loading design tables from Stat-Ease Design-Expert® software version 9 or 10.
- Added ‘Perten-DX’ import for loading data files in JCAMP-DX format created by Perten instruments.
- Added ‘Guided Wave NIRO JSON’ import for loading data files in JSON format created by the NIR-O™ Process Analyzer spectrometer by Guided Wave.
- Automatic tagging of spectral data is performed for some import functions.
- ‘Unscrambler’ import (for import of pre-X Unscrambler models) has been modified to support also test set and category validated models.
- ‘GRAMS’ import has been updated with v9.2 GRAMS components and the preview dialog performance has been improved.
Changes to PCA/PCR/PLSR methods
- NIPALS algorithm used in PCA, PCR, PLSR and OSC has been modified for faster convergence. Max. iterations can be set by user and convergence statistics are reported.
- New dialogs for PCA, PCR and PLSR for improved navigation.
- The previous ‘Warning limits’ and ‘Set Alarms’ functionalities in PCA, PCR and PLSR are replaced by optional ‘Outlier limits’ used to detect suspect data points in both calibration and prediction models.
- Option for calculating exact Q-residual limits in PCA, PCR and PLSR has been included.
- Option for allowing missing values in prediction has been added for PCR and PLSR. When unchecked, prediction results will usually be missing when prediction data are missing. Prediction results will be returned only if the variables in question do not influence the model,
- Change of model structure of PCA, PCR and PLSR to include category variables from raw data tables. This can be used for sample grouping and for recalculate when a category variable has been used to define cross-validation segments.
- In PCA, PCR and PLSR, the column header for all variables included in the original, non-preprocessed input data matrix is saved with the model. This is useful for mapping variables with instrument output in online applications.
- Options for Recalculate with/without test set samples have been added.
- Frozen calibration samples added to cross-validation setup.
- Rank testing included for PCA, PCR and PLSR. If rank of data are exhausted (to single precision), the number of components will be reduced accordingly. In some cases this can happen e.g. if model is created on data with predominantly additive effects in the first component. In this case most spectral transforms can be used to remove the additive effects and more components can then be found.
- The F-Residual Limit has changed compared to previous versions. It is now calculated based on calibration residuals rather than validation residuals. It follows that current limits are more conservative and old version models should therefore be recalculated in order to update the limits.
- A new sample grouping dialog with more options that is also simple to use has been added.
- Remember previous view in an open project. When clicking a model node in project navigator, the previous view is loaded rather than redrawing the defaults.
- A range of new plots added to full prediction models, including Contribution plots.
- Copy with headers allowed from numeric views of plots.
- ‘Spectra’ columnsets are selected by default in some spectra-oriented transforms.
- Dialogs for keep-out info have been modified with expand and collapse options.
- Row and Column header names are required to be present and unique. If not, they will be automatically be modified accordingly.
- An option to disable the ‘View Plots?’ dialog box when opening a project or performing modeling has been added (‘Tools—Options—General’).
- When duplicating (entire) matrices, pretreatment metadata are retained such auto-pretreatment options can be available for prediction at later stage. Also row and column ranges are retained in the duplicated matrices.
- ‘ASCII-MOD’ export was obsolete and it has been removed. (For some time it will be made available as a separate plugin for those who still use it.)
- Recalculate option has been restricted for Prediction models so that it is no longer allowed to change calibration model in the recalculate dialog. The matrix ‘X Original’ has be removed from prediction model node as a consequence.
- General enhancements and bugfixes.