How to “Compare calibration models”

Use appropriate tools
to fit and predict the two-way models.

For PARAFAC regression
fit a PARAFAC model to the calibration data and regress the dependent variable onto these scores. Calculate the scores of the validation data from the loadings of the PARAFAC model and use the regression coefficients for predicting the color.

Decide on
how/whether to center the spectral data prior to PARAFAC modeling and/or prior to regression.

Note that
since the PARAFAC model is not nested, each PARAFAC model has to be fitted separately.