Replication Data for: Pork adulteration detection in minced beef powders using near-infrared (NIR) spectroscopy

This dataset was generated to evaluate the potential of near-infrared (NIR) spectroscopy as a rapid and reliable method for detecting adulteration in minced meat products. It contains NIR spectra of powdered minced meat to detect pork-adulteration in minced beef. The samples consist of beef-pork mixtures in powdered form, prepared according to an experimental process that included heating at 40 °C for 20 hours, grinding, and sieving. Two types of samples were produced: lean mixtures (‘L’) and fat-rich mixtures (‘F’) containing 10% fat. Overall, this resulted in eight sample categories: BL (lean beef), 20L (lean beef with 20% pork adulteration), 50L (lean beef with 50% pork adulteration), 100L (lean pork), BF (fat-rich beef), 20F (fat-rich beef with 20% pork adulteration), 50F (fat-rich beef with 50% pork adulteration), and 100F (fat-rich pork). These eight sample categories were prepared in 10-fold (10 batches), representing 10 different sources of beef and pork.
Spectral measurements were collected with a Bruker MPA FT-NIR instrument over the wavelength range of 1125–2624 nm. All spectra are expressed in absorbance units as log(1/R). The dataset is organized into six blocks: (1) all spectra, (2) all Y block, (3) calibration (Cal) spectra, (4) calibration Y block, (5) test spectra, and (6) test Y block. The complete dataset, encompassing all batches, was divided into a calibration and a test set using the Kennard–Stone algorithm, while keeping samples from the same batch together. These data were used to develop and evaluate partial least squares discriminant analysis (PLS-DA) models for discriminating pork-containing samples from pure beef samples.