Statistical aggregation

What AreaCacao computes on close

When a tasting moves to Closed, the system runs an automatic statistical aggregation over the panel's responses.

Per sample and per attribute

• Mean: the average of the panel's scores.
• Standard deviation (SD): the spread. A high SD means the panel does not agree.
• Minimum and maximum: the extreme values.

Defects

• Defect percentage: for each defect in the protocol, what percentage of the panel flagged it.

Comparison across samples (when there are 2 or more)

• ANOVA (Analysis of Variance): a statistical test that detects whether the means across samples are statistically different.
• Tukey HSD: if the ANOVA returns p < 0.05, this post-hoc test identifies which pairs of samples differ significantly.

How to interpret the ANOVA

• p < 0.05 → there is a significant difference between at least two samples. Tukey HSD tells you which ones.
• p ≥ 0.05 → you cannot claim the samples differ. They may be the same, or the panel may not have detected the difference.

How to interpret Tukey HSD

When the ANOVA detects differences, Tukey HSD groups the samples into homogeneous groups and expresses them with letters (ᵃ, ᵇ, ᶜ): samples that share a letter do not differ significantly from one another. At a glance, you know which samples are indistinguishable to your panel and which ones stand out. This is the basis for practical decisions (for example, "batch A and batch C share a letter: they are interchangeable for our panel").

How to read it in the results table

In the tasting's Results section, the table of the descriptive report (or hedonic, or ranking with scales) presents this aggregation as follows:

• Each row is an attribute from the protocol.
• Each column is a sample, identified by its blind code.
• In each cell you will see the mean ± standard deviation of the panel for that attribute and that sample.
• The Tukey letters (ᵃ, ᵇ, ᶜ) appear as superscripts next to the mean: samples that share a letter do not differ significantly.
• The ANOVA column summarizes whether the attribute varies across samples, with a badge:
  • Difference (green) — there is a statistically significant difference.
  • No difference (gray) — no difference is detected.
  • n/a — there is not enough data for the test.

At the bottom of the table you will see this exact note:

Mean ± standard deviation per sample. Letters (ᵃ ᵇ ᶜ) indicate Tukey groups: samples sharing a letter do not differ significantly (α = 0.05).

Attributes with a significant difference between samples are also marked with an asterisk (*) in the spider chart.

Honest limitations

• ANOVA and Tukey HSD require at least 3–4 panelists to have statistical power. With smaller panels, the result is indicative only.
• They assume normality and homogeneity of variances (the usual assumptions). If your tasting has a very heterogeneous panel, consider external non-parametric analyses.
• The table shows aggregates per attribute and sample (mean ± SD), not the raw responses of each panelist one by one. There is also no CSV or Excel export from the results screen; the analysis is reviewed and interpreted within the tasting itself.