The art of converting complex chemical information into simple answers. Given a spectrometer recording 500 datapoints at different wavelengths, how do I find how much resin is in this plywood (fat/butter, octane/fuel,...).
Chemometrics largely depends on multivariate statistics to get the thing done. Although this might sound like cool tech, most chemist / spectroscopists would prefer to do without. A chemometric calibration can be a painstaking effort requiring 100 reference samples to be analysed in the lab. How long this calibration is going to hold is often anybody's guess. Calibrations on many agricultural products like wheat have to be redone every year. It is however the price to pay if you want to user easier sampling techniques, get faster results etc.