•Classical time-intensity is extended with MATI (multi-attribute time-intensity).•Data was collected with descriptive and consumer panels on taffy chews.•MATI offers concurrent evaluation of multiple ...intensity and hedonic attributes.•Some analytics are novel and avoid limitations in other time–intensity analytics.•MATI statistical methods offer panel and panelist performance assessment.
Taffy is a popular food form for delivery of functional ingredients but requires formulation that delivers acceptable flavor and texture throughout the entire product consumption experience. Because consumer-relevant flavor and texture changes occur throughout the mastication process for this type of product, it is useful to apply a sensory time–intensity methodology during chew-down for product optimization. Classical time–intensity methods are not efficient approaches for rapid development timelines as they are generally limited to single attributes per run. A multi-attribute time–intensity (MATI) application has been developed and applied to ‘pace’ respondents through multiple attributes and cycles within a run, thereby offering an efficient means to capture key flavor and texture attributes over time.
MATI is a natural extension of the classical time–intensity methodology, from one attribute to multiple attributes and from intensity attributes to both intensity and hedonic attributes. Many advanced statistical techniques can be used for analyzing the MATI data. The techniques include turning raw discrete data into smooth MATI curves; the bootstrap method for estimations of the parameters of MATI curves and the standard errors of the estimators; HANOVA, an adaptive analysis of variance for high-dimensional data, for comparisons of groups of MATI curves; and the intraclass correlation coefficient (ICC) and multivariate intraclass correlation coefficient (MICC) for assessing performance of a panel and panelists. A combination of MATI results from a trained sensory descriptive panel and consumers was used to deliver product category understanding, provide detailed understanding of what flavors and textures consumers enjoy and formulation optimization guidance.
Conventional drivers of liking analysis was extended with a time dimension into temporal drivers of liking (TDOL) based on functional data analysis methodology and non-additive models for ...multiple-attribute time-intensity (MATI) data. The non-additive models, which consider both direct effects and interaction effects of attributes to consumer overall liking, include Choquet integral and fuzzy measure in the multi-criteria decision-making, and linear regression based on variance decomposition. Dynamics of TDOL, i.e., the derivatives of the relative importance functional curves were also explored. Well-established R packages 'fda', 'kappalab' and 'relaimpo' were used in the paper for developing TDOL. Applied use of these methods shows that the relative importance of MATI curves offers insights for understanding the temporal aspects of consumer liking for fruit chews.
•Simulated Type I errors rates of ANOVA based on GLM and LM models for CATA data.•Simulated testing powers of ANOVA based on the GLM and LM models for CATA data.•Predicted precisions using GLM and LM ...models for CATA data.•The claim that logistic regressions violate Type I error rates is not generally true.•Liberty or conservativeness is not a criterion of validity or invalidity of a test.
Some discussions about statistical models used for analysis of CATA data appear in recent issues of this journal. This paper is a further discussion on the topic, following Bi and Kuesten (2022) (Food Quality and Preference, 95), related to Meyners and Hasted (2021, 2022) (Food Quality and Preference, 92, 95). This paper presents some statistical analyses for a real consumer CATA dataset using a generalized linear model (GLM) and a linear model (LM), respectively. The main objectives are to simulate Type I error levels of ANOVA; to simulate the empirical testing powers of ANOVA; and to compare testing results and predicted precisions based on different models. Meyners and Hasted (2022) claim that logistic regressions violate Type I error rates. The simulation results show that the claim is not generally true and suggest that violating Type I error rates is due to two-way ANOVA with a special type of Sums of Squares (Type2SS), not due to the GLM. Meyners and Hasted (2022) conclude that the GLM or logistic regression is invalid and flawed for analysis of CATA data. We disagree with the conclusion because we cannot find any convincing reasoning supporting the conclusion. It is true that the test of two-way ANOVA with Type2SS based on the GLM is more liberal than that with Type1SS and that based on the LM. Liberty and conservativeness are a characteristic of a test, not a criterion to judge validity or invalidity of a test. Meyners and Hasted (2022) advocate that GLM should be precluded for the use of CATA data. Our position is that the GLM deserves to be a standard and first selected practice, at least one of the useful and valid methods for analysis of CATA data. The advocacy of precluding the GLM for analysis of CATA data is unacceptable.
•GLM is more appropriate for testing product difference with CATA data.•Multiple comparisons based on GLM with a binomial distribution for CATA data.•Cochran’s Q test and normal approximation for ...CATA data have historical importance.•There is no sound reason nowadays to use these methods for CATA data analyses.
ANOVA and multiple comparisons based on generalized linear model (GLM) with a binomial distributionare proposed to be used for testing product differences with CATA data. The methods of Cochran’s Q test and ANOVA based on a normal distribution for CATA data analyses have only historical significance. There is no sound reason nowadays to insist on using these methods for CATA data analyses from both a theoretical and a practical perspective.
Demographics and psychographics are used to study the influence of different consumers on product effects in food development and testing. Demographics have a longer history and are routinely used in ...most research; psychographics are more recent, raising the question of whether they add to research on food products. The research presented here represents extensive exploratory data that demonstrate that both demographic measures and psychographic measures add to our understanding of consumer's liking ratings for nutrient supplements. The results are discussed in the context of broader research on a range of food products. In addition, the research reported here was conducted in four different countries, demonstrating many country effects. Finally, tests were conducted with users of the products, lapsed users of the product, and users of other nutrient supplements (non-users), and this led to many differences in product testing. These results further suggest that age and gender are not the only demographic variables to be studied, along with psychographic variables. The psychographic variables should be selected for a particular product category under investigation, as effects of specific psychographic measures vary for product categories. Specific variables do not fit all products for both demographics and psychographics.
The Two‐Out‐of‐Five method is one of the basic sensory discrimination methods. It is a special case of the ‘M+N’ method with M = 3 and N = 2. Three variants of the method are discussed in the paper: ...specified, unspecified, and unspecified with forgiveness. Analytical and simulation‐derived psychometric functions for all three variants of the Two‐Out‐of‐Five method are derived and produced. The performance of the method in both difference and similarity/equivalence tests are explored. It is shown that both difference and similarity/equivalence testing powers for the specified Two‐Out‐of‐Five are larger than those for the 2‐AFC and 3‐AFC. Both difference and similarity/equivalence testing powers for the unspecified Two‐Out‐of‐Five are larger than those for the Triangle and Duo‐trio. Both difference and similarity/equivalence testing powers for the unspecified with forgiveness Two‐Out‐of‐Five are larger than those for the conventional unspecified Two‐Out‐of‐Five and unspecified Tetrads. Tables and R codes are presented and provided for estimations of the probability of correct response, Pc, Thurstonian discriminal distance δ or d', and the B value for estimating the variance of d' for the three variants of the Two‐Out‐of‐Five method.
Practical Applications
The Two‐Out‐of‐Five method is a powerful sensory discrimination method. However, this method has not been used widely in sensory and consumer research, partly since the method has not been explored adequately in the literature. This paper explores the Two‐Out‐of‐Five method and shows that the method has potential for wide application. The Two‐Out‐of‐Five method involves a larger number of samples (5) and sensory fatigue may be of concern. Hence, the method may not be suitable for taste and smell tests with strong stimuli. The method is particularly suitable for manual and visual inspection testing where sensory fatigue is of a lesser concern.
•TURF (Total Unduplicated Reach and Frequency) is proposed for analyzing CATA data.•CATA examples from a women’s multi-vitamin/mineral gummy survey are shared.•R code ‘turfcata’ is provided and used ...with the R package ‘turfR’.
CATA (Check All That Apply) questions are one of the most popular approaches used widely in sensory and consumer fields. This paper proposes to apply the TURF (Total Unduplicated Reach and Frequency) technique to summarize and analyze CATA data. Numerical examples from a women’s multi-vitamin/mineral gummy survey conducted online recently in the US are provided to show the TURF analysis for CATA data. The R package ‘turfR’ was used in the analysis of the CATA data. An R code (‘turfcata’) was developed and is provided for the analysis using the R package ‘turfR’.
•Performance reliability of sensory descriptive panels is critical to sensory analysis.•Intraclass correlation coefficient (ICC) and Cronbach’s coefficient alpha can be used to measure performance ...reliability (PR).•Thurstonian d-prime and R-index can be used to measure Effect Size (ES).•Performance Reliability versus Effect Size (PR-ES) can be used as a framework for analysis of sensory descriptive panel data.•An historical protein powder case study is shared using the PR-ES framework.
Performance reliability of sensory descriptive panels is critical to sensory analysis. Treatment effect size (ES) of products is a main objective of sensory analysis. The intraclass correlation coefficient (ICC) and Cronbach’s coefficient alpha can be used to measure performance reliability, while Thurstonian d-prime and R-index can be used to measure ES. In this paper, an industry case study with a full historical database spanning 10 years of protein studies is shared. This case study uses the analysis framework of Performance Reliability versus Effect Size (PR-ES). R codes for the analysis are developed, used, and provided in the paper.
The paired A‐Not A with AB and BA pairs Bi, Jian; Kuesten, Carla
Journal of sensory studies,
February 2022, 2022-02-00, 20220201, Letnik:
37, Številka:
1
Journal Article
Recenzirano
This article explores the two‐interval, two‐alternative forced‐choice (2I2AFC) paradigm in signal detection theory (SDT), which can also be called the paired A‐Not A with AB and BA pairs in the ...sensory and consumer fields. The article describes how to estimate the parameters and their variances from the psychometric function of the method. Tables and R codes were developed and are provided for the estimations. The performance of the method is evaluated. It shows that the paired A‐Not A with AB and BA pairs is more effective and powerful than the conventional A‐Not A and the A‐Not A with reminder (A‐Not AR) tests. An extension of the method, that is, the ratings of the paired A‐Not A with AB and BA pairs is also explored. A facial shine case study applying this method is provided.
Practical applications
As one of the paired versions of sensory discrimination methods, the paired A‐Not A with AB and BA pairs has wide potential applications to detect and measure sensory difference, especially to detect and measure before‐after treatment effects of products. The method can be used to detect both product effects and serving order effects.
•The 4I2AFC is a paired 2-AFC with AB and BA pairs.•The 4I2AFC can be used to detect small changes of stimuli.•The 4I2AFC can be used to identify direction of stimuli changes.•The 4I2AFC is more ...powerful than the tetrads in both difference and similarity testing.•The 4I2AFC is particularly suitable for visual or manual evaluations with less sensory fatigue.
The four-interval, two-alternative forced-choice (4I2AFC) is a paired version of the 2-AFC with AB and BA pairs as two alternatives, where A is a signal or a stronger stimulus and B is a noise or a weaker stimulus. The task of the subject in 4I2AFC is to select the pair (AB) with decreasing stimuli change. The 4I2AFC is a powerful sensory discrimination forced-choice method to detect small, directional changes of stimuli particularly suitable for visual or manual evaluations with less sensory fatigue. This paper explores the 4I2AFC method and includes an applied illustrative case study for measuring facial skin smoothness visually using the method. An analytical psychometric function is derived. The performances of the 4I2AFC in both difference testing and similarity/equivalence testing are compared with those of the conventional sensory discrimination forced-choice methods. The 4I2AFC is theoretically more powerful than any of the published conventional methods including the tetrads.