Guillermo Hough

Principal component analysis in sensory analysis: covariance or correlation matrix?

When principal component analysis (PCA) is applied to descriptive analysis, the input data is a sample (rows) by descriptor (columns) matrix, usually formed from the mean values over assessors. This data matrix is the input to the PCA procedure of statistical softwares, which presents the option of performing PCA on either the covariance matrix (cov-PCA) or the correlation matrix (corr-PCA), both derived from the data matrix. A non-comprehensive survey of papers where PCA was used to analyze sensory descriptive data, showed that out of a total of 52 papers, 22 used corr-PCA, seven used cov-PCA and 23 did not say which PCA method they used. PCA of three real sensory data sets, showed how the results may change by either using cov-PCA or corr-PCA. Cov-PCA should be used in most cases as the sensory scales are the same for all attributes. Corr-PCA should only be used when there is a very good reason for doing so, rather than the reverse.

Partial least squares correlations between sensory and instrumental measurements of flavor and texture for Reggianito grating cheese

Abstract Visual, manual and oral texture descriptors for Reggianito grating cheese were correlated with parameters obtained from an instrumental compression test. Aroma and flavor descriptors were correlated with organic acid concentrations. Linear partial least squares regression was used for the multivariate correlations and proved to be a useful tool. Measurements were done between 30 and 290 days of ripening time. Strain at breaking point was the instrumental texture parameter which best correlated with sensory descriptors: negatively wth visual, manual and oral fracturability; and positively with manual elasticity. Sensory hardness was related to instrumental hardness although the correlation coefficients were not high. Total flavor and aroma intensity were well correlated with propionic acid, suggesting this acid could be a flavor indicator for this type of cheese. Uric and orotic acids changed during ripening yet remained below sensory threshold values.

Perceptions of genetically modified foods by consumers in Argentina

Abstract Perception of genetically modified foods (GMF) by consumers in Argentina was investigated using the repertory grid method in conjunction with generalized Procrustes analysis. The following factors were considered: type of genetic modification (microbial, plant or animal), rationale for modification (nutritional, sensory or economic), labeling or not labeling as genetically modified, controls (local or international) and associated risks (health or environment). 20 men and 20 women were recruited from a typical Argentine town. Differences in perception between men and women were not significant. Participants used a greater number of negative constructs (“risky to health” and “changing the environment”) than positive constructs (“beneficial”, “improvements in nourishment”). They want these foods to be regulated and labeled as such, showing distrust in Argentine health organizations. Some participants considered GMF as “beneficial” if produced to improve their nutritional value, but “superficial” and “whimsical” if produced to improve sensory properties. Participants separated statements according to the rationale for modification, but not according to whether the modification was on an animal, plant or microorganism.

Descriptive analysis and external preference mapping of powdered chocolate milk

Abstract Appearance, texture and flavor descriptors were developed for powdered chocolate milk. The influence of cocoa and gum concentrations on these descriptors was studied using stepwise multiple regression. Out of a total of 23 descriptors, four were non-significant. For the significant descriptors, the percentage variance explained ranged from 65 to 93%, with an average of 82%. Visual viscosity and oral thickness were correlated with instrumental viscosity. Principal component analysis showed appearance/texture was explained by four principal components, whilst aroma/flavor was one-dimensional and depended on cocoa concentration alone. The circular ideal point model was chosen to map consumers on the first two appearance/texture principal components. For aroma/flavor, consumers were mapped on a single dimension using an ideal point model. For both appearance/texture and aroma/flavor, adults and children had similar averages but the ideal points for children were more spread out on the sample map.

Preference mapping of cracker-type biscuits

Abstract The sensory profile and acceptability of nine commercial samples of cracker type biscuits were measured. The descriptors used by a trained sensory panel that best separated samples were: appearance: color, quantity, uniformity, distribution, size and breakage of hills, hole related descriptors and ease of brand reading; aroma and flavor: total intensity, sweet, salty, cheesy, margarine, flour and toasted; texture: rough and powdery surface, crispness, noise, hardness, oiliness and greasiness. A total of 360 regular cracker consumers (1/3 youths and 2/3 adults) from three different cities in Argentina measured appearance, texture, flavor and overall acceptance. On average, adults scored crackers higher in acceptability than youths; likewise consumers from the smaller inland towns scored higher than consumers from Buenos Aires, the largest city in Argentina. Average overall acceptance scores for the nine samples ranged from 62 to 76 on a 0–100 scale with a least significant difference (LSD) of 4. Seven, nine, nine and seven consumer clusters were found for appearance, texture, flavor and overall acceptance, respectively. For each attribute there was one cluster representing between 10 and 23% of consumers who scored all samples high. Extended preference mapping showed the sensory descriptors of crackers which drove the preference of some of the consumer clusters.

Sensory optimization of a powdered chocolate milk formula

Abstract Four experiments were carried out to optimize the formulation of a powdered chocolate milk. The first of these compared facial vs. verbal-numeric self-administered scales when measuring sensory acceptability of different attributes among 11-year-old children. Both scales were equivalent. In a second experiment, magnitude estimation was performed to analyze sensory viscosity as a function of gum concentration, and chocolate flavor as a function of cocoa concentration. Stevens Law exponents were: 0.37 for viscosity and 0.78 for chocolate flavor. Third, by using a minimum bias method, the optimum level of sugar concentration resulted in 5.0% for adults and 6.0% for children. In the fourth experiment gum and cocoa concentration were optimized using response surface methodology. Based on Stevens Law exponents, levels were varied geometrically. Contour plots for appearance, texture, flavor and overall preference are presented for children and adults. No unique formula would satisfy both age groups for all attributes.

How Argentine consumers understand the Spanish translation of the 9-point hedonic scale

Abstract The objective of this work was to study how ambiguous the meaning of the Spanish translation of the 9-point hedonic scale was for Argentine consumers. Three versions of the 9-point hedonic scale were tested: a literal translation of the original English version, a liberal translation of this same scale and the liberal translation of a scale phrased for children (e.g. where ‘like extremely’ was replaced by ‘super good’). Children, adolescents and adults from two cities of Argentina, making up a total of 288 subjects, performed the tests. These consisted in ranking the phrases from each one of the three scales from ‘like least’ to ‘like most’. In most cases the scale phrased for children performed best. Using this scale approximately 30% of the subjects ranked the translated phrases differently in relation to the English version, that is inverted the order of two or more phrases. The results from the present work show that the use of the 9-point hedonic scale translated to languages different to English should be done with caution.

Survival analysis: A consumer-friendly method to estimate the optimum sucrose level in probiotic petit suisse

For years, just-about-right (JAR) scales have been among the most used techniques to obtain sensory information about consumer perception, but recently, some researchers have harshly criticized the technique. The present study aimed to apply survival analysis to estimate the optimum sucrose concentration in probiotic petit suisse cheese and compare the survival analysis to JAR scales to verify which technique more accurately predicted the optimum sucrose concentration according to consumer acceptability. Two panels of consumers (total=170) performed affective tests to determine the optimal concentration of sucrose in probiotic petit suisse using 2 different methods of analysis: JAR scales (n=85) and survival analysis (n=85). Then an acceptance test was conducted using naïve consumers (n=100) between 18 and 60 yr old, with 2 samples of petit suisse, one with the ideal sucrose determined by JAR scales and the other with the ideal sucrose content determined by survival analysis, to determine which formulation was in accordance with consumer acceptability. The results indicate that the 2 sensory methods were equally effective in predicting the optimum sucrose level in probiotic petit suisse cheese, and no significant differences were detected in any of the characteristics related to liking evaluated. However, survival analysis has important advantages over the JAR scales. Survival analysis has shown the potential to be an advantageous tool for dairy companies because it was able to accurately predict the optimum sucrose content in a consumer-friendly way and was also practical for researchers because experimental sensory work is simpler and has been shown to be more cost effective than JAR scales without losses of consumer acceptability.

Sensory profiling during ripening of reggianito grating cheese, using both traditional ripening and in plastic wrapping

Abstract Reggianito grating cheese was manufactured and ripened under two conditions: traditionally and wrapped in air-tight plastic. A sensory profile to describe this type of cheese was developed with a total of 52 descriptors covering attributes of external appearance; color; visual, manual and oral texture; aroma and flavor. Principal component analysis was found adequate to reduce dimensionality and follow attribute variation versus ripening time. Wrapping this type of cheese in plastic bags modified its appearance and texture, and to a lesser extent its flavor. For grating, consumers preferred samples ripened traditionally. For oral preference, scores were equivalent.

Note: Relationships of Consumer Acceptability and Sensory Attributes of Yerba Mate (Ilex Paraguariensis St. Hilarie) using Preference Mapping:

A sensory profile and acceptability test of nine commercial brands of yerba mate (Ilex paraguariensis St. Hilarie) were carried out. The descriptors used by a trained sensory panel that best separated brands were: stick and leaf size, stick and leaf size uniformity, quantity ofsticks and quantity of dust, included in the appearance of the dry yerba; sediment, turbidity and brown colour that corresponded to appearance of the infusion; and initial impact, acid, humid, smoke, paper, chemical, green, toasted and residual, for aroma and flavour. A total of 106 consumers (half men, half women) tested overall acceptance of the 9 brands in a home location test. Average overall acceptance scores for the 9 samples ranged from 42 to 63 on a 0–100 non-structured scale with an LSD of 8. Gender did not influence consumer acceptability forthe different yerba mate brands. Based on overall acceptance consumers were clustered into five groups. Extended internal preference mapping showed the descriptors of yerba mate that drovethe preference of some of the consumers clusters. Both appearance and aroma/flavour descriptors were correlated to preference dimensions.