Extraction dataset and analysis code for a source-verified quantitative synthesis of > banana-residue-derived supercapacitor electrodes

Description

This deposit contains the complete extraction dataset, analysis code, intermediate results, and figure-generation pipeline supporting a quantitative research synthesis of supercapacitor electrodes derived from banana biomass residues (peel, pseudostem, bract, leaf, and fibre). The dataset comprises 28 electrode conditions extracted from 16 peer-reviewed primary studies, each independently verified as indexed in Scopus or Web of Science, with every reported value confirmed against the publisher's record of the primary source. The unit of analysis is the electrode condition rather than the paper, because individual studies commonly report several conditions; the 28 rows are therefore **not** 28 independent studies, and any re-analysis must account for the clustering of conditions within studies. Recorded fields cover the banana fraction used, conversion route, activation agent and temperature, heteroatom doping, BET surface area, pore volume and pore-size class, electrolyte identity and concentration, electrode configuration, measurement current density or scan rate, and the reported outcomes of specific capacitance, energy density, power density, capacitance retention and cycle count. Two provenance fields record, for every value, whether it was read from the publisher's full text or an open abstract, and the indexing-verification status of the source journal. A separate sheet documents every excluded record with its exclusion reason, including values that were removed during verification because they could not be confirmed in the primary source. The accompanying code reproduces every statistic, table, and figure reported in the associated article from the raw extraction, in ten documented stages: descriptive and bivariate analysis; a leave-one-out machine-learning layer; a robustness suite covering distributional form, rank-based re-estimation, bootstrap intervals, false-discovery-rate correction and variance inflation; PRISMA and effect-size figures; subgroup and influence diagnostics; a conditional meta-analytic apparatus reporting fixed-effect and random-effects pooling, Cochran's Q, I-squared, H-squared, tau-squared, prediction intervals and the Egger and Begg tests; a reproducible six-domain risk-of-bias assessment; measurement-basis stratification with cluster-robust and wild cluster bootstrap inference; a statistical power and feasibility analysis; and an assessment of compliance with existing published reporting guidelines for electrochemical capacitors. Reported specific capacitance is not commensurable across records: values span two- and three-electrode configurations, measurement rates across more than four orders of magnitude, and electrolytes with different stable voltage windows, and no study states the mass basis on which gravimetric quantities are reckoned.

Steps to reproduce

Python 3.10 or later. Approximately 200 MB free disk space. No network access required: the pipeline reads only `data/Master_Dataset.xlsx`.```bash python3 -m venv venv source venv/bin/activate # Windows: venv\Scripts\activate pip install -r requirements.txt ``` The analysis was executed and verified on Python 3.10.12 with numpy 2.2.6, pandas 2.3.3, scipy 1.15.3, statsmodels 0.14.6, scikit-learn 1.7.2, shap 0.49.1, matplotlib 3.10.9 and openpyxl 3.1.5, on Linux x86-64 with glibc 2.35. Exact versions are pinned in `environment_versions.json`. ```bash bash run_all.sh ``` 7.3 Or run stages individually Run in numerical order; later stages read earlier outputs. ```bash cd code python3 01_statistical_analysis.py # -> results/statistical_results.json, Figures 5-9 python3 02_ml_layer.py # -> results/ml_results.json, Figure 10 python3 03_robustness_analysis.py # -> results/robustness_results.json python3 04_figures.py # -> Figures 1, 11, graphical abstract python3 05_meta_figures.py # -> results/meta_figures_results.json, Figures 2, 12, 13 python3 06_meta_analysis.py # -> results/meta_analysis_results.json, Figures 14, 15 python3 07_risk_of_bias.py # -> results/risk_of_bias_results.json, Figures 16, 17 python3 08_stratified_synthesis.py # -> results/stratified_results.json, Figures 3, 4 python3 09_power_and_feasibility.py # -> results/power_results.json python3 10_guideline_compliance.py # -> results/guideline_compliance.json, Figure 18 ``` Each script prints its full results to stdout as well as writing JSON. ### 7.4 Confirm you obtained the same answer ```bash sha256sum -c CHECKSUMS.sha256 ``` The `data/` and `results/` checksums must match exactly. **Figure checksums may differ on a different machine even when the underlying numbers are identical**, because PDF and PNG writers embed creation timestamps and subset fonts differently. Compare the JSON files in `results/` rather than the image files when checking reproduction; the images are byte-unstable but pixel-identical. 7.5 Determinism All stochastic components are seeded, so repeated runs on the same input give identical output. | Component | Seed | Location | |---|---|---| | Random forest regressor | 42 | `02_ml_layer.py` | | Histogram gradient boosting regressor | 42 | `02_ml_layer.py` | | Bootstrap CI on cross-validated R-squared (2,000 resamples) | 42 | `02_ml_layer.py` | | Bootstrap CI on median capacitance (10,000 resamples) | 2026 | `03_robustness_analysis.py` | | Wild cluster bootstrap (9,999 Rademacher replications) | 42 | `08_stratified_synthesis.py` | | Cluster bootstrap on the intraclass correlation (400 replications) | 42 | `09_power_and_feasibility.py` | | Jitter in the configuration strata plot | 7 | `08_stratified_synthesis.py` | Leave-one-out cross-validation is deterministic by construction and uses no seed.

Institutions

• Capiz State University

  Western Visayas, Roxas City

Categories

Licence