Effects of packaging system and storage temperature on the physicochemical and microbiological stability of smoked catfish and rabbit
Description
This dataset accompanies the study £Effects of packaging system and storage temperature on the physicochemical and microbiological stability of smoked catfish and rabbit". It provides the package-level experimental data and reproducible statistical workflow used to evaluate changes in smoked catfish (*Clarias gariepinus*) and rabbit (*Oryctolagus cuniculus*) during 90 days of storage. The storage experiment compared three packaging systems—vacuum PA/PE pouches, low-density polyethylene (LDPE) pouches and grease-proof paper—under ambient conditions (approximately 25–30 °C) and refrigeration (4–5 °C). Post-packaging assessments were conducted at 30, 60 and 90 days. Two separately prepared packages were analysed for each package × product × storage × time combination, providing 72 package-level observations per outcome for factorial inference. Day-0 measurements represent common product starting conditions and are retained for descriptive purposes but excluded from package- and storage-related factorial analyses. The measured outcomes are water activity, peroxide value, pH, mesophilic aerobic bacterial count, coliform count, yeast count and mould count. The repository includes the long-format package-level dataset, data dictionary, annotated R analysis script, primary factorial-model results, complete-factorial sensitivity results, model diagnostics, comparison of lower-order effects across model specifications and validation documentation. The primary statistical analysis evaluates package, observed product category, storage condition, time and all two-way interactions. A complete four-factor model containing all higher-order interactions is provided as a sensitivity analysis. The deposited workflow also documents the transformations, treatment of microbiological non-detects, Benjamini–Hochberg false-discovery-rate correction and model diagnostics used in the study. The dataset is intended to support transparent reanalysis, statistical reproducibility and further methodological investigation of packaging and storage behaviour in smoked animal-source foods. It should not be interpreted as a commercial shelf-life or food-safety validation dataset. Processing-run identifiers were not retained in the package-level storage dataset; therefore, between-smoking-run variation cannot be estimated, and the catfish–rabbit factor represents an observed product-category comparison rather than an independently replicated species effect.
Files
Steps to reproduce
Download and extract all repository files, preserving the folder structure. The main files are `data/storage_stability_package_level_data.csv`, `code/storage_stability_analysis.R`, and the reference outputs in the `outputs` folders. Run the analysis in R from the repository root using: `Rscript code/storage_stability_analysis.R data/storage_stability_package_level_data.csv reproduced_outputs` Base R is sufficient. The dataset contains package-level observations for three packaging systems (Vacuum, LDPE, Paper), two product categories (Catfish, Rabbit), two storage conditions (Ambient, Refrigerated), and three inferential assessment times (30, 60 and 90 days). Two separate packages were analysed per treatment-time cell, giving 72 post-baseline observations per outcome. Day-0 observations are common product baselines used only for descriptive summaries and figures and must be excluded from factorial inference. Analyse water activity and pH on their recorded scales. Transform peroxide value as `log10(PV + 1)`. Microbial counts are analysed as `log10 CFU/g`; bacterial non-detects are assigned 25 CFU/g and yeast/mould non-detects 5 CFU/g before transformation. Duplicate laboratory readings are technical replicates and are averaged within package. The primary model is: `response ~ (Package + Meat + Storage + Time)^2` using sum-to-zero contrasts. It includes four main effects and all six two-way interactions and retains 52 residual degrees of freedom. Benjamini-Hochberg correction is applied globally across the 70 primary omnibus tests. The complete sensitivity model is: `response ~ Package * Meat * Storage * Time` and includes all interactions through fourth order, retaining 36 pure-error residual degrees of freedom. Benjamini-Hochberg correction is applied separately across the 105 complete-model tests. Compare generated results with `primary_anova_main_two_way.csv`, `complete_factorial_all_terms.csv`, `lower_order_model_comparison.csv`, and `model_diagnostics.csv`. Small rounding differences are acceptable. The experimental unit is the individual package. Treatment means and technical replicates must not be analysed as additional observations. Processing-run identifiers were not retained, so between-smoking-run variation cannot be estimated, and the Catfish-versus-Rabbit factor should be interpreted as an observed product-category comparison rather than a fully replicated species effect.
Institutions
- CSIR - Animal Research InstituteGreater Accra, Accra
- Kwame Nkrumah University of Science and TechnologyAshanti, Kumasi