The ``working_pls
'' node in SciCraft which contains the Partial
Least Squares Regression (PLSR) method is an Octave (Matlab) m-file
which has the following syntax:
[T,W,Yval,B,b0,MSEC,MSEP,P,Pvalues,aopt] = working_pls(X,Y,Amax)
The input arguments are:
: The input data matrix. Each row is an object and each
column contains the different variables (features)
: The response matrix with the dependen data
the maximum number of PLSR components which will be
tested for. The optimal number of PLSR factors are determined by
cross validation in the ``working_pls
'' node.
The output arguments from the node are:
: Matrix of PCA scores (the coordinates in the new
coordinate system defined by the principal components)
: Matrix of loading weights
-
cross validation validatin and
calibration data
-
: The regression coefficient vector/matrix and
the offset. These are used for predictions of new samples.
-
Cross validation errors for calibration
and validation.
The loadings matrix
-
Significance p-values for each variable from
Jackknife analysis of the cross validation
segments[2,3,1].
-
The optimal number of PLSR factors
Bjørn Alsberg
2005-02-18