Weight by Colour
Distribution per colour
Box = quartiles, points = every individual M&M
Bag composition
Count of each colour measured
Weight density by colour
Weight distribution per colour (Gaussian KDE)
The Tasters
Who measured more (by count)
Number of M&M's recorded per taster
Who measured more (by weight)
Total grams recorded per taster
Calibration check
Mean weight: taster × colour
Hall of Fame
Statistics by Colour
When colour started to matter
Mean weight ± 95% CI
Which colours differ?
Each cell compares two colours (Welch t-test, Holm-adjusted). Green means they differ: * p<0.05 · ** p<0.01 · *** p<0.001. Hover a cell for the verdict.
| Colour | n | Mean (g) | SD | SEM | Median | Min | Max |
|---|
Methods
M&M's (n = …) were obtained from commercially available bags of unverified provenance. Colour was assigned by eye under ambient laboratory lighting. Each specimen was weighed exactly once on a laboratory balance and then destroyed by ingestion. The authors consider the reason self-evident.
The effect of colour on weight was tested with a one-way ANOVA, followed by pairwise Welch t-tests with Holm correction. Effect size is eta-squared (η² = SSbetween / SStotal), the share of weight variance explained by colour: … here. Colour proportions were tested against a uniform mix with a chi-square goodness-of-fit test (expected count N / k per colour, k = colours observed, k − 1 degrees of freedom). All numbers are recomputed from the spreadsheet on each load.
Author contributions: P.T., M.G., J.V., C.B.-R. and G.B. performed the experiments. G.B. analysed the data and drafted the manuscript. All authors consumed the samples and approved the final version.
Funding: This work was generously supported by the ERC, the Emergency Reserve of Chocolate.
Ethics statement: no specimen survived the study.
The authors have declared no competing interest.