Using results from plsregress to determine unknowns

Hey everyone,

I have a data set of Raman spectra that I have used PLSregress to perform a calibration. My calibrants are an 18x2351 X-matrix of Raman spectra and a 18x2 Y-matrix of known concentrations for 2 compounds. Using plsregress I can get a beta matrix (2352x2), BUT I don't understand how to use this beta matrix to predict the concentration of an unknown sample. I have 5 samples I want to validate the model with, but I can't figure out what to do next! (Note: my linear algebra is a little weak)

Thanks!

Colin

Answer 1

[Darshan Ramakant Bhat]

You can just add a bias column (column of ones) to your X_validate matrix and multiply with beta. Sample code can be as follows

Y_validate = [ones(5,1),X_validate]*beta;

The result matrix Y_validate will be of dimension 5X2.

I hope this answer will help to solve your problem.

Thanks and regards

Darshan Bhat

[Colin Young]

That makes a lot of sense. Thanks!