Research Data – A multidimensional framework for characterizing rum terroir: development and application in the French West Indies
Description
This dataset contains the research data underlying the study “A multidimensional framework for characterizing rum terroir: development and application in the French West Indies”. The dataset covers rum-producing units in Guadeloupe and Martinique and includes anonymized producer information for 28 production units (24 agricultural-rum and 4 industrial-rum units). The analytical datasets comprise 47 agricultural rum products (25 white and 22 aged), originating from Guadeloupe and Martinique. Physicochemical and GC-FID analyses were performed in triplicate. The GC-FID dataset includes measurements of 22 volatile compounds, with 15 compounds retained for multivariate analyses based on complete data availability. The data package includes anonymized producer information, raw physicochemical and GC-FID measurements, a data dictionary, PERMANOVA/PERMDISP statistical results, a README file and two Python scripts supporting the reproducibility of the multivariate analyses. All production units and rum samples are identified using anonymized identifiers. The dataset provides a transparent and reproducible basis for investigating the physicochemical and volatile composition of agricultural rums from the French West Indies. These data are made available to promote transparency, reproducibility and further research on rum terroir characterization and its contribution to rum typicity in the French West Indies.
Files
Steps to reproduce
The producer-provided information was anonymized and structured at the production-unit level. Missing or non-reported information was retained according to the study coding framework. Physicochemical measurements were performed in triplicate for each rum sample, and raw replicate measurements are provided. GC-FID analyses were also performed in triplicate for each rum sample, and raw measurements for 22 volatile compounds are provided. For multivariate analyses, replicate measurements were averaged for each rum sample. Fifteen volatile compounds with complete data across all 47 agricultural rums were retained, while seven compounds with missing chromatographic peaks were excluded. No missing-value imputation was performed. White (n = 25) and aged (n = 22) rums were analysed separately following z-score standardization within each category. Principal component analysis (PCA) was performed on the standardized data. Euclidean distances were used for multivariate comparisons. Differences in volatile composition between territorial groups were assessed using PERMANOVA, based on Gower-centred squared Euclidean distances, with 9,999 unrestricted permutations. Homogeneity of multivariate dispersion was assessed using PERMDISP based on distances to group centroids, with 9,999 unrestricted permutations. The provided Python scripts reproduce the PCA, PERMANOVA and PERMDISP analyses. The accompanying data dictionary and README document the variables, data structure and analytical workflow. These datasets support the reproduction of the statistical analyses reported in the associated study.
Institutions
- Université des AntillesGuadeloupe, Pointe-à-Pitre