Handbook of Biomedical Optics

Raman, SERS, and FTIR Spectroscopy 243

only c

t

in each. Somewhat surprisingly, this is sucient to pre-

dict the concentration c

t

in future specimens robustly, as long

as enough samples are scanned (J must exceed N, and should in

practice be at least three times greater). In the absence of noise

sources and other nonidealities, implicit calibration methods

are just as accurate as OLS. Implicit calibration is in fact usu-

ally more robust, because it naturally characterizes real-world

sources of additional spectral variation, such as baseline dri

and residual uorescence, just as it would a genuine Raman

component.

e workhorse for implicit ...