What mdatools can do?
The package includes classes and functions for analysis, preprocessing and plotting data and results. So far the following methods for analysis are implemented:
- Principal Component Analysis (PCA)
- Soft Independent Modelling of Class Analogy (SIMCA), including data driven approach (DD-SIMCA)
- Partial Least Squares regression (PLS) with calculation of VIP scores and Selectivity ratio
- Partial Least Squares Discriminant Analysis (PLS-DA)
- Randomization test for PLS regression models
- Interval PLS for variable selection
- Multivariate curve resolution using the purity approach
- Multivariate curve resolution using the constrained alternating least squares
- Procrustes cross-validation for PCA
Preprocessing methods include:
- Mean centering, standardization and autoscaling
- Savitzky-Golay filter for smoothing and derivatives
- Standard Normal Variate for removing scatter and global intensity effect from spectral data
- Mutliplicative Scatter Correction for the same issue
- Normalization of spectra to unit area, unit length, unit sum, unit area under given range.
- Baseline correction with asymmetric least squares
- Kubelka-Munk transformation
- Element wise transformations (
sqrt, power, etc.)
Besides that, some extensions for the basic R plotting functionality have been also implemented and allow to do the following:
- Color grouping of objects with automatic color legend bar.
- Plot for several groups of objects with automatically calculated axes limits and plot legend.
- Three built-in color schemes — one is based on Colorbrewer and the other two are jet and grayscale.
- Very easy-to-use possibility to apply any user defined color scheme.
- Possibility to show horizontal and vertical lines on the plot with automatically adjusted axes limits.
- Possibility to extend plotting functionality by using some attributes for datasets.
?mdatools and next chapters for more details.