calculates the eigenvectors and the parameters for tuning External Parameter Orthogonalization (EPO) with the classical approach based on the means for each influence factor
res = epo(x_ed,classes_indiv,classes_perturb,xcal,ycal,split,lv,(centering))
a matrix (n1 x q) or a Div structure from an experimental design
a conjunctive vector or a disjunctive matrix identifying the different samples of x_ed for which observations were acquired
a conjunctive vector or a disjunctive matrix identifying the different levels of the detrimental influence; their number is nbr_perturb
calibration dataset; a matrix of spectra (n x q) and a vector of reference values (n x 1) or Div structures
for the cross validation; a scalar representing a number of contiguous blocks, or a vector identifying each observation to a block
number of latent variables for the PLS regression
centred=1 (by default); not centred=0
the D matrix containing only detrimental information
res.d_matrix.d is a matrix of dimensions ((nbr_perturb-1) x q)
eigenvectors of d_matrix
res.eigenvec.d is a matrix of dimensions (q x (nbr_perturb-1))
eigenvalues of d_matrix in percent
res.d_ev_pcent.d is a vector of dimensions ((nbr_perturb-1) x 1)
Wilks lambda
res.wilks.d is a vector of dimensions (nbr_perturb x 1)
rmsecv for several dimensions of EPO and several dimensions of PLSR
res.rmsecv.d is a matrix of dimensions (lv x nbr_perturb)
PLS models obtained after an EPO correction using 0/1/2/...(nbr_perturb - 1) eigenvectors from res.eigenvect.d
res.pls_models is a list of dimensions (nbr_perturb)
help pls for more information about the fields of res.pls_models