Hello everyone,

I have a question related to the model development using the Learn R PCR model component in Pipeline Pilot. Given a set of training data, descriptors and maximum number of latent variables I understand that the component outputs the normal and adjusted cross-validated Root Mean Squared Error of Prediction (RMSEP) and % variance explained as a function of the latent variables and also builds a model for further use. Nonetheless the manual does not explain very clearly how the model is produced, or in other words, how the training is performed.

What I do not understand is how the optimal number of latent variables is chosen for actually building the model and made available for making new predictions in the "Learned Properties" folder. 

Put differently, how one can know how many latent variables the deployed model has retained, these being anywhere between 1 and the maximum number of latent variables chosen by the user.

Thank you!