Frank Westad

Variable selection in near infrared spectroscopy based on significance testing in partial least squares regression

A jack-knife based method for variable selection in partial least squares regression is presented. The method is based on significance tests of model parameters, in this paper applied to regression coefficients. The method is tested on a near infrared (NIR) spectral data set recorded on beer samples, correlated to extract concentration and compared to other methods with known merit. The results show that the jack-knife based variable selection performs as well or better than other variable selection methods do. Furthermore, results show that the method is robust towards various cross-validation schemes (the number of segments and how they are chosen).

Gender specific preferences and attitudes towards meat

Abstract Recently, red meat avoidance has shown an increase in the industrialised countries, especially among young female consumers. Sensory factors as bloodiness in meat, difficulties coping with eating a fellow animal, and private body concern appear as the main reasons for red meat exclusion. The study addressed whether sensory attributes in meat are linked to attitudes and beliefs about meat. Based on previous studies, the expectation that red meat is linked to dislike and negative attitudes among young females was tested. The study used a quantitative approach, applying both a quantitative sensory profiling with trained panellists and a consumer study with a convenience sample. The trained sensory panel evaluated 22 sensory attributes of five meats, ranging from red (beef) to white (chicken) meat varieties. Comparable samples of the same meat varieties were served in randomised order to 206 young consumers, males and females between the ages of 14 and 30 years, in a blind preference test. Beliefs and attitudes towards meat-eating, and desired change in consumption frequencies of flesh products were also collected. Consumers preferred the white meat (chicken) to the red meats. The mean hedonic rating of meat decreased progressively as the meat increased in red colour intensity and typical meat flavours, and this was particularly evident for females. Females displayed, in contrast to males, significantly lower mean hedonic scores for the reddest meat varieties, i.e. ostrich, lamb and beef. Males displayed, compared with females, also a significantly higher attitudinal support for “pro-red meat” statements. The results were strengthened by significantly higher desired increase in consumption frequency of beef among male consumers. The link between consumer and product was established and revealed a close relationship between specific sensory attributes of meats and consumer attitudes towards meat. For example, sensory attributes related to white meat were correlated with negative attitudes towards red meat. The hypothesis that dislike of red meat varieties is more prevalent among females was supported.

Regression of a data matrix on descriptors of both its rows and of its columns via latent variables: L-PLSR

Abstract A new approach is described, for extracting and visualising structures in a data matrix Y in light of additional information BOTH about the ROWS in Y, given in matrix X, AND about the COLUMNS in Y, given in matrix Z. The three matrices Z–Y–X may be envisioned as an “L-shape”; X(I×K) and Z(J×L) share no matrix size dimension, but are connected via Y(I×J). A few linear combinations (components) are extracted from X and from Z, and their interactions are used for bi-linear modelling of Y, as well as for bi-linear modelling of X and Z themselves. The components are defined by singular value decomposition (SVD) of X′YZ. Two versions of the L-PLSR are described—using one single SVD for all components, or component-wise SVDs after deflation. The method is applied to the analysis of consumer liking data Y of six products assessed by 125 persons, in light of 10 other product descriptors X and 15 other person descriptors Z. Its performance is also checked on artificial data.

Incorporating chemical band-assignment in near infrared spectroscopy regression models

In this paper, we present an approach for incorporating chemical band assignment information in regression models between spectra and constituents. It is shown how the matrices in this L-shaped data structure can be combined and give direct information of the relationships between theoretical chemical band assignment, spectral wavelengths and the responses. The chosen application is NIR spectroscopic measurements of canola seeds. Variable selection based on partial least squares regression using jack-knifing within a cross-model validation (CMV) framework is applied for removing non-relevant spectral regions. Extended multiplicative scatter correction was applied as a spectral pre-treatment to remove physical scatter effects in the spectra. The results show a high degree of correspondence between the objectively found wavelength bands from CMV and the reported chemical interpretation found in the literature.

Analysis of designed experiments by stabilised PLS Regression and jack-knifing

Abstract Pragmatical, visually oriented methods for assessing and optimising bi-linear regression models are described, and applied to PLS Regression (PLSR) analysis of multi-response data from controlled experiments. The paper outlines some ways to stabilise the PLSR method to extend its range of applicability to the analysis of effects in designed experiments. Two ways of passifying unreliable variables are shown. A method for estimating the reliability of the cross-validated prediction error RMSEP is demonstrated. Some recently developed jack-knifing extensions are illustrated, for estimating the reliability of the linear and bi-linear model parameter estimates. The paper illustrates how the obtained PLSR “significance” probabilities are similar to those from conventional factorial ANOVA, but the PLSR is shown to give important additional overview plots of the main relevant structures in the multi-response data. The study is part of an ongoing effort to establish a cognitively simple and versatile approach to multivariate data analysis, with reliability assessment based on the data at hand, and with little need for abstract distribution theory [H. Martens, M. Martens, Multivariate Analysis of Quality. An Introduction, Wiley, Chichester, UK, 2001].

Variable selection in PCA in sensory descriptive and consumer data

This paper presents a general method for identifying significant variables in multivariate models. The methodology is applied on principal component analysis (PCA) of sensory descriptive and consumer data. The method is based on uncertainty estimates from cross-validation/jack-knifing, where the importance of model validation is emphasised. Students t-tests based on the loadings and their estimated standard uncertainties are used to calculate significance on each variable for each component. Two data sets are used to demonstrate how this aids the data-analyst in interpreting loading plots by indicating degree of significance for each variable in the plot. The usefulness of correlation loadings to visualise correlation structures between variables is also demonstrated.

Detection of Parasites in Cod Fillets by Using SIMCA Classification in Multispectral Images in the Visible and NIR Region

The presence of parasitic nematodes in fillets of commercially important fish species has been a serious quality problem for the fishing industry for several decades. Various approaches have been tried to develop an efficient method to detect the parasites, but so far the only reasonable solution is manual inspection and trimming of each fish fillet on a candling table. In this study we have investigated how multispectral imaging in combination with SIMCA classification can be used for automatic detection of parasites. The results indicate that the spectral characteristics of nematodes differ sufficiently from those of fish flesh to allow one to obtain fairly good classifications. The method is able to detect parasites at depths down to about 6 mm into the fish muscle. The method shows promising results, but further studies are required to verify feasibility for the fish industry.

Cross validation and uncertainty estimates in independent component analysis

A data analysis tool, known as independent component analysis (ICA), is the main focus of this paper. The theory of ICA is briefly reviewed, and the underlying statistical assumptions and a practical algorithm are described. This paper introduces cross validation/jack-knifing and significance tests to ICA. Jack-knifing is applied to estimate uncertainties for the ICA loadings, which also serve as a basis for significance tests. These tests are shown to improve ICA performance, indicating how many components are mixed in the observed data, and also which parts of the extracted sources that contain significant information. We address the issue of stability for the ICA model through uncertainty plots. The ICA performance is compared to principal component analysis (PCA) for two selected applications, a simulated experiment and a real world application.

Shift and intensity modeling in spectroscopy—general concept and applications

Abstract Multivariate analysis and calibration in spectroscopy have been widely used in many fields of applications in recent years. The number of components to use in such applications for sufficiently describing the spectra is dependent on the stability of positions of the peaks of interest. Unwanted shifts give rise to a more complex model, i.e., more components. Various kinds of shift correction can be applied as preprocessing tools to cope with these effects. In this paper we describe a general concept of how to model intensity (peak height) and position as separate phenomena. We described methods and theory from image modeling, and apply them on one-dimensional signals such as traditional Raman spectroscopy.

Colour of ground beef as influenced by raw materials, addition of sodium chloride and low oxygen packaging

The study aimed at examining the effects of freezing of raw materials, holding time for fresh raw materials post mortem and addition of 0.5-1.0% NaCl on the colour of ground beef under low oxygen (O2) modified atmosphere storage. The samples were exposed to 0.1-3.0% O2 at 4°C for up to 10 days, and analysed for O2 concentrations, instrumental and visual colour. Residual O2 in the headspace of the packages oxidizes myoglobin and discolours the meat. Meat may have the ability to scavenge residual O2, and ground beef differs from intact muscles by having a much higher capacity for O2 consumption. In this experiment, the use of frozen/thawed raw materials and addition of NaCl both decreased the rate of O2 consumption and increased discolouration. Using raw materials from 2 days rather than 7 days post mortem greatly increased the rate of removal of O2 and improved redness. In low O2 packaging, ground beef preferably should be stored for at least 2 days in an atmosphere with less than 0.1% residual O2 to produce a purple pigment predominantly consisting of deoxymyoglobin.