Performs Orthogonal Partial Least-Squares analysis on matrix X.
Utilises the mjrMainO2PLS function written by Mattias Rantalainen.
model = csm_orth_pls( spectra, Y );
model = csm_orth_pls( spectra, Y, 'num_pred_comp', num_pred_comp, 'num_Y_orth_comp', num_Y_orth_comp, 'num_cv_rounds',num_cv_rounds, 'cvType',cvType, 'scale_type',scale_type, 'cv_frac',cv_frac, 'model_type',model_type, 'cv_pred',cv_pred, 'orth_pls_type',orth_pls_type, 'largeBlockSize',largeBlockSize);
| Variable | Type | Default Value | Description |
|---|---|---|---|
| *spectra | csm_spectra | None | csm_spectra object containing spectral matrix. |
| *Y | m*1 | None | Matrix of predictors - Orthogonal components (For discriminant analysis this is a vector of 0/1's to define class) |
| num_pred_comp | 1*1 | 1 | Number of predictive components |
| num_Y_orth_comp | 1*1 | 3 | Number of Y-orthogonal components (OC in X). The number of components in num_pred_comp+numYorthComp should be kept to a minimum to prevent overfitting |
| num_cv_rounds | 1*1 | 7 | Number of cross-validation rounds. Often set to zero |
| scale_type | str | 'mc' | 'uv' for unit variance scaling, 'pa' for pareto, 'mc' for mean centred, 'none' for no scaling |
| model_type | str | 'da' | 'da' for discriminant analysis, 're' for regression. if 'da', sensitivity and specificity will be calculated |
| Variable | Type | Description |
|---|---|---|
| csm_orth_pls | csm_wrapper | csm_wrapper with some stored inputs, the outputs and metadata. |
| csm_orth_pls.output.da | str | if model is of model_type='da' (discriminant analysis |
| csm_orth_pls.output.o2pls_model | struct | the full o2pls model with all it's parameters |
| csm_orth_pls.output.cv | struct | cross-validation results, such as Q^2 values, |
| csm_orth_pls.output.release | str | version information |
O2PLS
Copyright Imperial College London 2019