Data analyses of a multiple‐samples sensory ranking test and its duplicated test: A review

Abstract

A multiple‐samples ranking test is a quick, simple, and useful tool to assess differences in preference or sensory attribute intensity among multiple products. In this test, each panelist evaluates and ranks a complete set of samples once, generating one vector of multiple dependent data. To fit this ordinal dependency, the nonparametric Friedman's test is required. Duplicated ranking tests require a statistical analysis method that accounts for an additional level of dependency between duplicates. The study of duplicated ranking testing has not received much attention until recently. This review discusses historical development of methods and statistical analysis of sensory ranking tests, leading to current practices, alternative procedures including duplicated ranking testing, some factors that induce errors, and statistical considerations for the duplicated multiple‐samples ranking test. Practical applications A multiple‐samples sensory ranking test has been used to compare multiple samples on the basis of ranks of preference or attribute intensity. Most of the applications of sensory ranking tests include early product sorting or screening in a product development process, studies in which children and elderly are target subjects, and descriptive panel screening. In some cases, to reduce the number of subjects, time and cost, duplicated ranking tests are performed and data are analyzed using the nonparametric Friedman's test, not taking into consideration additional dependency between duplicates. Duplicated ranking testing can be beneficial if data analysis is properly handled, which is discussed in this review article.

Functional and textural properties of vegetable‐fibre enriched yoghurt

“…; Vidal et al . ; Carabante and Prinyawiwatkul ) or innovative methods based on consumer perception as CATA, pivot profile (Esmerino et al . ; Torres et al .…”

“…For many years, sensory attribute analysis has been performed using descriptive analysis with trained assessors. Methodological research is needed to support improvement in the discrimination and reliability for product characterisation by consumers using projective mapping and principal component analysis (Janiaski et al 2016;Vidal et al 2016;Carabante and Prinyawiwatkul 2018) or innovative methods based on consumer perception as CATA, pivot profile Torres et al 2017).…”

Comparing Friedman versus Mack–Skillings data analyses on duplicated rank data: a case of visual color intensity

“…This computation provides multiple comparisons based on experiment‐wise error rates. More information about how to compute the Mack–Skillings test and post hoc multiple comparisons is available …”

“…In multiple‐samples ranking tests, n panelists receive a set of k samples to rank according to an attribute intensity or preference. Non‐parametric tests and tables of critical values are widely chosen over traditional analysis of variance (ANOVA) for the analysis of rank data . The Friedman rank‐based nonparametric test, asymptotically following a chi‐squared distribution, and the associated P value, provide a measure of the degree of overall differences.…”

“…Non-parametric tests and tables of critical values are widely chosen over traditional analysis of variance (ANOVA) for the analysis of rank data. 2 The Friedman 3 rank-based nonparametric test, asymptotically following a chi-squared distribution, 4,5 and the associated P value, provide a measure of the degree of overall differences. Among non-parametric methods, the Friedman test, 3 detailed by Hollander et al, 5 is widely recommended for the randomized block design (RBD) without a panelist * sample interaction when used in sensory ranking.…”