Example of PCR, interpretation of calibration equations

1
Example of PCR, interpretation of calibration
equations

• NIR analysis of 3 constituents
• Casein
• Lactate
• Glucose
• Scatter corrected (eliminates unwanted effects)

2
 
Exp. design
PC176.5 PC223.1 SUM99.6
All objects were measured by NIR (many hundred
var. in X) The NIR spectra were submitted to
PCA Number of variables much larger than the
number of objects
3
 
Principal components in regression
Loadings
Weighted regression coefficients Glucose 2.10
and -0.31 Lactate -0.65 and 1.65 Two components
explain gt 99 of y Cross-validation
Mean centred glucose spectrum
Lactate-casein spectrum
4
Conclusions

• The NIR data matrix can be adequately modelled by
  a linear Beers law model (linear function of
  constituents)
• Good predictions are obtained
• Glucose is closely related to the first component
  and then responsible for more spectral
  variability than the other two constituents
• Which wavelengths are affected by the different
  constituents. Verified by comparing with spectral
  patterns of the constituents
• No outliers, strengthening the confidence of the
  results.
• The interpretation of scores, X-loadings and
  y-loadings fit well together