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

Abstract

© 2019 Society of Chemical Industry BACKGROUND: In duplicated ranking tests, panelists either rank duplicates separately (2SS) or jointly in a single session (1SS). This study compared data analyses of duplicated yellow color intensity rank data using Friedman versus Mack–Skillings (M–S) tests. Panelists (n = 75) ranked an orange juice set twice – a similar-samples set (100%, 95% versus 90%); samples other than the 100% juice were prepared by dilution with water. Rank sum data were obtained from (a) intermediate ranks from jointly re-ranked scores of 2SS for each panelist, and (b) joint rank data of all panelists from the two replications in 1SS. Both (a) and (b) were analyzed using the M–S test. The median rank data (c) for each panelist from two replications were analyzed using the Friedman test. RESULTS: Comparing M–S with the Friedman tests, the former generally produced higher test statistics and lower P-values than the latter. However, when considering the pattern of post hoc pairwise significant differences, both tests yielded similar conclusions. CONCLUSION: This study demonstrated that, in a duplicated ranking test with three samples that were very similar in color, separating the two replications into two complete individual ranking tests or serving sessions (2SS) may prevent sensitivity loss due to fatigue that is otherwise experienced when evaluating all samples together in a single session (1SS). We expected to find the M–S test to be more sensitive than the Friedman test; however, this hypothesis was not supported by the post hoc (Tukey's honest significant difference (HSD)) multiple comparison test results under the specific test conditions in this study. © 2019 Society of Chemical Industry.