Visualizing indirect correlations when predicting fatty acid composition from near infrared spectroscopy
measurements

Problems concerning covariance among independent variables are well understood and dealt with by inverse regression methods like partial least squares regression. However, covariance between dependent variables has only received minor attention. Biological samples are often complex mixtures of multiple covarying compounds. During multivariate calibration, analyte predictions may be mediated through relationships with interfering compounds, which implies that the calibration model is not providing a direct link between the multivariate measurements and the analyte of interest. This compromises robustness and validity of the calibration model-important aspects when applying the model to future samples and data sets. This study discusses issues of calibration modeling when strong covariance structures exist among the analyte of interest and interfering compounds.We propose a projection-based method to diagnose whether indirect covariance structures dominate the calibration model. The proposed method is tested on a two-constituent Beer's law system consisting of 20 aqueous samples with covarying amounts of fructose (analyte of interest) and riboflavin (interfering compound). Transmission measurements are obtained on all samples in the visual and near-infrared wavelength ranges. Riboflavin has strong absorption in the visual region, whereas fructose exclusively absorbs in the near-infrared region. Hence, predictions of fructose concentrations, obtained from the visual wavelength range only, are fully mediated through riboflavin, whereas fructose predictions obtained from the near-infrared wavelength range may be obtained independent of riboflavin.

show abstract

“…This ensures the model to, exclusively, fit on the chemical basis related to c 1 and the relationship between c 1 and

{\overset{truê}{bold-italicc}}_{1}

would be given by

s_{1}^{T} \overset{b}{true}

(Equation ). For a good and direct model, the

(‖‖, \overset{truê}{bold-italicb})

is scaled so

s_{1}^{T} \overset{b}{true}

is close to 1 5,19 . These observations are also evident from Equations and .…”

Section: Theorymentioning

confidence: 78%

Diagnosing indirect relationships in multivariate calibration models

Eskildsen

Engelsen

Dankel

et al. 2021

Journal of Chemometrics

View full text Add to dashboard Cite

show abstract

“…Since moderate positive correlations between protein and hydroxyproline content were seen in both data sets, it was important to assure that the calibrations for collagen did not rely on the total protein content. One way of studying this is by investigating the correlations between the predicted values for protein and hydroxyproline contents, respectively (Eskildsen, Naes, Wold, Afseth, & Engelsen, 2019). Thus, PLSR models for protein were obtained, and the correlation coefficients between predicted protein and predicted hydroxyproline was calculated (r= 0.64 and r= 0.78 for beef and poultry by-products, respectively).…”

Section: Regression Analysismentioning

confidence: 99%

Raman, near-infrared and fluorescence spectroscopy for determination of collagen content in ground meat and poultry by-products

Monago-Maraña

Wold

Rødbotten

et al. 2021

LWT

View full text Add to dashboard Cite

Raman, near-infrared and fluorescence spectroscopy were evaluated for determination of collagen content in ground meat. Two sample sets were used (i.e. ground beef and ground poultry byproducts), and collagen concentrations (measured as hydroxyproline) varied in the ranges 0.1-3.3% in the beef samples and 0.4-1.5% in the poultry samples. Similar validation results for hydroxyproline were obtained for NIRS (R 2 = 0.82 and RMSECV = 0.11%) and Raman (R 2 = 0.81 and RMSECV = 0.11%) for the poultry samples. For the beef samples, NIRS obtained slightly less accurate results (R 2 = 0.89, RMSECV= 0.25%) compared to Raman (R 2 = 0.94, RMSECV= 0.19%), most likely due to less representative sampling. Fluorescence spectroscopy gave higher prediction errors (RMSECV= 0.50% and 0.13% for beef and poultry, respectively). This shows that Raman spectroscopy employing a scanning approach for representative sampling is a potential tool for on-line determination of collagen in meat.

show abstract