Aggregate data and analysis code for: Consumer preference and flavour signatures of Korean monofloral and creamed honeys (blind tasting, n = 322)

This deposit contains the complete aggregate data and analysis code underlying the short communication "Consumer preference and flavour signatures of Korean monofloral and creamed honeys: an exploratory blind tasting with 322 visitors at an agricultural technology fair." A two-day blind tasting was conducted on 18-19 June 2026 at the Agricultural Technology Fair (Osong Convention Center, Republic of Korea). A total of 322 visitors (day 1: n = 163; day 2: n = 159) tasted eight honeys produced by young Korean beekeepers - acacia (Robinia pseudoacacia, served on both days), wildflower, creamed, Hovenia dulcis, cherry blossom (Prunus spp.), Styrax japonicus, and chestnut (Castanea crenata) - and selected one favourite honey and the flavour term best representing each sample from a 14-term list. Contents (single archive): - data/ : favourite-honey votes per sample per day, overall flavour-sticker counts (14 terms, N = 761), top-three flavour counts per sample-day unit, and a table of all reported values with verification tolerances - scripts/ : 01_reproduce_all.py (recomputes all statistics), 02_verify.py (checks computed against reported values), 03_figures.py (regenerates Figures 1 and 2 at 300 dpi) - DATA_DICTIONARY.md, README.md Running the two scripts reproduces every statistic reported in the manuscript and verifies each one: the expected output is "99 MATCH, 0 MISMATCH, 0 MISSING". Analyses include chi-square goodness-of-fit tests with Cohen's w, Bonferroni-corrected adjusted standardised residuals, Wilson confidence intervals, two-sample proportion and Fisher exact tests, flavour-signature tests (lift, odds ratios, continuity-corrected 2x2 chi-square), and exploratory Spearman correlations. Note: data are aggregate counts only; no individual-level records exist. Multiple flavour selections were permitted, so flavour-count tests are approximate. Preference was measured by single choice and therefore reflects relative preference rather than acceptability. Sample codes and spatial arrangement were not counterbalanced. Tested with Python 3.10+ (numpy, scipy, pandas, matplotlib).