Study-level dataset of matched hydrogen-storage effects in MOF-carbon/biochar hybrid systems

Description

This dataset contains study-level data underlying an exploratory quantitative synthesis of hydrogen-storage changes in metal-organic framework (MOF)-carbon/biochar hybrids and surface-modified derivatives. The evidence corpus comprises 20 unique full-text experimental studies. Twelve studies provide matched direct adsorptive comparisons measured at the same temperature and pressure and constitute the primary quantitative dataset, while eight mechanistically related MOF-derived carbon/hydride studies are retained separately for contextual interpretation. The primary dataset includes study identity, material system, temperature, pressure, comparator and modified hydrogen-storage capacities, ratio of means (RoM), lnRoM, relative change, carbon-origin class, modification strategy, comparator class, and source location. Supporting files include a data dictionary, study references, contextual-study data, and Python analysis code that reproduces the study-level geometric mean RoM, exact sign test, nonparametric bootstrap interval, subgroup summaries, leave-one-out analysis, and chronological cumulative synthesis. The dataset is intended to support transparent reuse, verification, re-analysis, and future extension.

Steps to reproduce

Download all files associated with this dataset and place 02_Matched_Hydrogen_Storage_Dataset.csv and 04_Analysis_Code.py in the same working directory. Install Python and the required packages: pandas, numpy, and scipy. Run the analysis script using: python 04_Analysis_Code.py. The script reads the study-level matched-comparison dataset, verifies the required variables and positive capacity values, and calculates the ratio of means (RoM) and natural-log ratio of means (lnRoM). The script performs the exact two-sided binomial sign test for directional consistency and calculates the unweighted geometric mean RoM. A nonparametric study-level bootstrap is performed using 200,000 resamples with random seed 42 to estimate the descriptive 95% bootstrap interval. The script also produces descriptive subgroup summaries, leave-one-out analysis, and chronological cumulative synthesis. The reference results are: 12 matched contrasts; 11/12 positive effects; exact two-sided sign-test p = 0.0063; geometric mean RoM = 1.446718; study-bootstrap 95% interval = 1.256183–1.671036; and leave-one-out geometric-mean range = 1.375454–1.519316. The supplied data dictionary and README file describe all variables, classification rules, and dataset structure. No missing study-specific standard deviations are imputed and no inverse-variance weighting is applied.

Institutions

• Bandung Institute of Technology

  West Java, Bandung

Categories

Licence