The function make calibration models ralating 2 types of data. Communly it is
used to relate component concentrations or properties to absorbance spectra of
a known set of reference samples. It is based on STD.ASTM E 1655-97, a standard
practices for infrared, Multivariate, quantitative analysis. 					

The function pcrpls make 2 types of models: PCR and PLS. It needs 2 matrixes:
One whith the spectra (X) and other whith the properties that you want to model
(the reference values Y). The correct sintax of the function is “pcrpls(X,Y)”.
First the function have some pre-treatments available for the spectra. Than the
function separate the samples in two groups: the validation set and de
calibration set. Than  in a loop the function makes 2 models one PCR and one
PLS, and make an analysis of the model and an outlier statistic based on the
Mahalanobis Distance and delete the samples that are considered outliers. The
loop continue to calculate models until the function can not find any more
outliers. In the process the function shows two grafics for each model. One
with the error of prediction for the validation samples and another whith the
real values of the properties, and the values calculated by the function. For
each model the function shows on the screen the minimum number of PRESS (the
sum of the squared errors of prediction for the validation samples), the number
of principal components of the model, the number of samples in calibration set,
the number of samples in the validation set, the t value to show if the
validation entimates show a statistically significant bias (the t value must be
less than the critical t value), the number of validation samples that falls
whithin the confidence limits (desireble to be more than 95%), the standard
error of calibration (SEC), and the standard error of validation (SEV).  For
each model the function save the important variables in an archive. The
variables are:

Mbeta (matrix of conversion)- PCR,
Mbeta1(matrix of conversion) - PLS; 
Xstd - standard deviaton of each point of the spectra for the calibration set 
Xm - Mean of each point values. 
Ym mean value of the reference values.
Normaliza – indicates if the spectra data were normalized
Autoescala – indicates if the spectra data were autoscaled
Centra – indicate if the spectra data were mean-centered
Resultadospls – For PLS – contains the results showed on the screen. 
Resultado – For PCR – contains the results showed on the screen.
Numdaamostra – This variable contains the number of the samples that were
inclused on the model 
Numdaamostraval – This variable contains the number of the samples that were
not excluded from the validation set.
out - This variable contains the number of the samples that were excluded from
the calibration set.
out1 – This variable contains the number of the samples that were excluded from
the validaton set.
Compara – PCR -Contains the real and the calculated values of the property
modeled.
Comparapls – PLS -Contains the real and the calculated values of the property
modeled.
After to calculate and choose the best of the models calculated for the
function pcrpls. The function preve can be used to predict the property that
was modelated. You will have to load the variables of the best model and the
spectra of the samples you need to know the results. To predict the property of
the spectra X you just write “preve(X)”.