Data and Python code for the analysis of dilution and feeding-tube caliber effects on transfer performance and nutrient recovery of a blenderized enteral formulation

This repository contains the data, analysis code, documentation, and generated outputs associated with the study entitled “Effects of Dilution and Feeding-Tube Caliber on Transfer Performance and Nutrient Recovery of a Blenderized Enteral Formulation.” The study evaluated a standardized blenderized enteral formulation using a balanced (4\times4) factorial design comprising four formulation concentrations (25%, 50%, 75%, and 100% (m/v)) and four feeding-tube conditions (4, 5, 8, and 14 Fr). Three independently prepared formulations were evaluated for each concentration–caliber combination, resulting in 48 experimental units. Each transfer procedure began with 200 g of formulation and was observed for a maximum of 30 min. The archived datasets include the original transfer records, bromatological results, and a linked experimental-unit dataset containing formulation concentration, tube caliber, preparation number, recovered mass, observed transfer time, and the total energy, macronutrient, fiber, and mineral contents associated with each recovered sample. The Python workflow reproduces the analyses reported in the manuscript, including: calculation of recovered mass percentage and mass-transfer rate; calculation of energy- and protein-recovery rates; calculation of nutrient density in the recovered fraction; descriptive summaries for the 16 experimental combinations; factorial models with concentration, tube caliber, and their interaction; heteroskedasticity-consistent HC3 inference; residual-permutation tests; aligned-rank sensitivity analyses; Holm-adjusted simple-effect comparisons; leave-one-preparation-out influence analyses; stratified non-parametric bootstrap analysis; bootstrap ranking stability; exact enumeration of the multi-objective non-dominated set; and generation of the interaction figures included in the manuscript. The bootstrap and permutation analyses were performed using a fixed random seed of 20260620. The default workflow uses 10,000 bootstrap iterations and 9,999 residual permutations. The repository contains: original transfer and bromatological spreadsheets; a processed procedure-level analysis dataset; a data dictionary; the complete Python analysis script; pinned software requirements; a README with execution instructions; statistical result tables; bootstrap and Pareto-analysis outputs; and publication-ready figures. All analyses were conducted in Python 3.13 using NumPy, SciPy, statsmodels, Patsy, and Matplotlib. The study was conducted entirely in vitro and the repository contains no patient-level or personally identifiable data. Keywords Blenderized tube feeding; enteral nutrition; feeding-tube caliber; dilution; transfer performance; nutrient recovery; bromatological composition; factorial design; bootstrap; multi-objective analysis; Python; reproducible research. Recommended resource type Dataset or Software, depending on whether the Zenodo record is intended to emphasize the experimental data or the reproducible analytical workflow. For a combined archive containing both, Dataset is usually the most suitable primary resource type, with the software described in the metadata.