Microwave dielectric spectra of fish products (processed prawns, cod, haddock and saithe and mixtures of cod and other food materials), pork and chicken have been measured and the data analyzed using the statistical method of Principal Component Analysis (PCA). The principal components were registered against various compositional variables, (water, protein, phosphorous and salt) and the calibration validated using the method of internal cross-validation.
It was found that dielectric spectra measurement or control and transformed by PCA could be used to predict certain composition variables for which this is true are liquid uptake or added water, protein and salt. Unfortunately the amount of polyphosphate added was only well predicted in cases where salt had not been added.
An unexpected finding was that the protein content could be determined with reasonable accuracy, often less than 1%. This also has implications for the measurement of added water since one of the current approaches is to measure protein (from nitrogen content) and relate it to added water from knowledge of the protein content of untreated samples.
The Principal Components were also used as discriminators for the separation of the samples into groups, e.g. added water or no added water, polyphosphate or no polyphosphate. PCA demonstrated its power in this discriminant analysis by enabling prior treatment of samples to be determined. It was also found that this applied to mixtures of fish and other food materials.
These prior treatments are listed below:
- Water added
- Polyphosphate added
-Salt added
-Fresh or frozen (and mixtures containing cod)
-Quality (type fish)
The last of these are surprising and require further investigation. The quality referred to is arbitrary in the sense that it refers only to the suitability of a given type or fish for a certain process.