StickForStats

Bridging the gap between complex statistical methods and biological interpretation

Python Django React Statistics Data Visualization
stickforstats.com Try PCA Prototype GitHub
A sample falls into a histogram under its fitted normal curve, then re-forms as a Q–Q plot that leaves the outliers off the line

Advanced Statistics for Researchers

StickForStats is an open-source web platform that makes statistical assumption checking the default step before every analysis, not an optional extra. Its Guardian engine has eight validators (normality, variance homogeneity, independence, outliers, sample size, modality, linearity, homoscedasticity) and runs the relevant ones before a test executes. It reports the result alongside its assumption checks or, when a critical assumption fails, reroutes to an appropriate nonparametric alternative with a documented decision trail. A companion module re-checks the statistical claims in uploaded manuscripts.

Built to counter the assumption violations that drive irreproducible results, StickForStats runs on Django REST Framework and React with PostgreSQL, is MIT-licensed, and has been validated against SciPy and R. In the preprint, re-analysing a 91-sample synovial-sarcoma RNA-seq study (GSE271517) with Guardian moved 90.6% of 27,221 genes to a rank-based test and changed the differential-expression verdict for 553 of them. The platform is live (public beta) at stickforstats.com, the code is on GitHub, and a Python SDK and command-line interface are on PyPI (pip install stickforstats).

Current Status: Preprint on bioRxiv (2026), DOI: 10.64898/2026.06.15.732278; in peer review at BMC Bioinformatics. First and corresponding author.

Assumption Validation

Guardian checks a test's assumptions before it runs and reroutes to a nonparametric alternative on critical violations

Experimental Design

Power and sample-size analysis, with an optional 50-decimal-digit precision mode

Interactive Visualizations

Interactive charts, including forest and volcano plots

Transparent Analysis

Reports assumption checks alongside every result and exports analyses as R or Python code

Platform Features

Assumption Validation (Guardian)

Before a test executes, Guardian runs the relevant subset of its eight validators in parallel and computes a composite confidence score. Critical violations reroute the analysis to an appropriate nonparametric alternative; an AI Statistical Advisor helps with test selection.

Educational Resources

Five curated biological example datasets with documented analysis vignettes. Early interactive Streamlit prototypes, such as the Confidence Intervals Explorer, explore statistical concepts through visualizations and simulations.

Power and Sample Size

Power and sample-size analysis for study design, with an optional 50-decimal-digit precision mode, plus effect sizes with confidence intervals.

Interactive Visualizations

Interactive charts built with Recharts, including forest plots for meta-analysis and volcano plots for differential expression.

Transparent Analysis

Reports assumption checks alongside every result, exports analyses as R or Python code, and offers Python and R SDKs for programmatic access.

Manuscript Review

Parses PDF, LaTeX and DOCX manuscripts, extracts their statistical claims and checks each one for internal consistency, with discipline profiles for CONSORT, STROBE, ICH-E9 and JARS-Quant.

Available Modules

Data Exploration & Visualization

  • Correlation Analysis (Pearson r, with Spearman's rho as the alternative when normality or linearity fails)
  • Interactive forest and volcano plots
  • Principal Component Analysis (PCA), an early Streamlit prototype
  • Probability Distributions explorer, an early Streamlit prototype
Try PCA Prototype

Hypothesis Testing

  • Tests checked by Guardian before they run (t-test, ANOVA, regression, chi-square)
  • Recommended Alternatives (Mann-Whitney U, Kruskal-Wallis, Spearman's rho, Welch's t-test, Welch's ANOVA)
  • Multiple Testing Correction (eight FWER and FDR methods, including Bonferroni, Holm-Bonferroni and Benjamini-Hochberg)
  • Power Analysis for sample size calculation
  • Effect sizes with parametric, bootstrap and noncentral confidence intervals

Biomedical Analyses

  • Per-gene differential expression with assumption checks and Benjamini-Hochberg FDR correction
  • Survival Analysis (Kaplan-Meier, Cox proportional hazards, log-rank test)
  • Meta-analysis (fixed and random effects, heterogeneity, Egger's publication-bias test)
  • Causal inference (propensity-score matching, inverse-probability weighting, difference-in-differences, mediation)

Educational Tools

  • Confidence Intervals Explorer, an early Streamlit prototype
  • Five curated biological example datasets with documented analysis vignettes
Try CI Explorer

Try the Early Streamlit Prototypes

StickForStats began as standalone Streamlit apps before moving to Django and React. Three of those early interactive prototypes are still online. They sleep when idle, so the first load may take a minute.

PCA Prototype

An early Principal Component Analysis module that lets researchers upload datasets, perform PCA with automatic preprocessing, and visualize results with customizable 2D/3D plots.

Try PCA Prototype

Probability Distributions

An early prototype for exploring common statistical distributions interactively: see how parameters affect distribution shapes, fit distributions to your data, and calculate probabilities and critical values.

Try Distributions Prototype

Confidence Intervals Explorer

An early prototype for understanding confidence intervals through interactive simulations: visualize how sample size affects precision, and compare frequentist and Bayesian approaches.

Try CI Explorer

Development Timeline

Apr 2024

Project Inception

Initial concept development and prototype design of the StickForStats platform.

Jun 2024

Core Modules Development

Development of key modules including PCA and Probability Distributions, released as standalone demos.

Jul-Aug 2024

Integration Phase

Combining individual modules into a unified platform with consistent UI/UX and shared data processing pipeline.

2025

Best Poster Award

EMBO Conference 2025, CCMB Hyderabad, for the poster “StickForStats: Publish Research with Confidence”. First author and presenter. View the poster (PDF, 2.4 MB)

2026

Preprint and Peer Review

Preprint posted on bioRxiv (DOI: 10.64898/2026.06.15.732278); in peer review at BMC Bioinformatics.

Who Benefits from StickForStats?

Experimental Biologists

Researchers conducting wet-lab experiments who need to analyze their data without extensive statistics training. StickForStats helps with experimental design, sample size calculation, and appropriate statistical analysis of results.

Example: A molecular biologist comparing gene expression changes across multiple treatment conditions can use StickForStats to check each test's assumptions, switch to a nonparametric alternative where they fail, and correct for multiple testing.

Clinical Researchers

Medical professionals conducting clinical studies who need robust statistical methods for patient data. StickForStats provides tools for survival analysis, meta-analysis, causal inference, and manuscript review against CONSORT and ICH-E9.

Example: A clinical researcher studying drug efficacy can use StickForStats to perform power analysis for study design, adjust for confounding with methods such as propensity-score matching, and check the manuscript's statistics before submission.

Graduate Students

Students in biology, medicine, and related fields who need to analyze their thesis data but lack formal statistical training. StickForStats serves as both an analysis tool and an educational resource.

Example: A PhD student can see which assumptions each test needs and whether their data meet them, work through the documented analysis vignettes, and export the analysis as R or Python code.

Technology Stack

Frontend

  • React 18: User interface
  • Material-UI 5: UI components
  • Recharts: Interactive visualizations, including forest and volcano plots
  • jStat: Client-side computations

Backend

  • Python 3.11, Django 4.2 and Django REST Framework: Application layer with Guardian integration
  • NumPy/SciPy: Array operations and statistical functions
  • statsmodels: Regression diagnostics and GLMs
  • lifelines: Survival analysis
  • scikit-learn: Machine learning utilities
  • mpmath: Optional 50-decimal-digit precision mode
  • Celery with Redis: Long-running analyses

Deployment

  • Docker and Docker Compose: Containerized deployment of the full stack
  • PostgreSQL 15 and Redis: Relational storage, caching and task queues
  • GitHub Actions: CI running more than 1,500 automated tests on every commit
  • PyPI: Python client SDK and command-line interface
  • Streamlit Cloud: Hosting the early standalone prototypes

Project Vision

StickForStats aims to democratize advanced statistical analysis by making complex methods accessible to researchers of all backgrounds. Our vision includes:

  • Reducing the statistical knowledge barrier by checking assumptions automatically and guiding users to an appropriate analysis
  • Improving reproducibility in research by providing standardized analysis workflows and transparent code
  • Building statistical literacy in the scientific community through integrated learning resources
  • Creating a platform for collaboration where researchers can share and build upon analytical approaches
  • Supporting open science principles by making the entire platform open-source under the MIT license

We believe that by making advanced statistical tools more accessible, we can help researchers focus on their scientific questions rather than struggling with statistical methods, ultimately accelerating discovery and improving research quality.

Get Involved

StickForStats is an open-source project, and we welcome contributions from the community:

Contribute Code

Help develop new modules, improve existing functionality, or fix bugs. We welcome contributions from developers of all skill levels.

GitHub Repository

Beta Testing

Use the public beta at stickforstats.com and send feedback on functionality, usability, and features that would benefit your research.

Send Feedback

Documentation

Help improve tutorials, documentation, and educational resources to make the platform more accessible.

Contact Us

Ready to Transform Your Statistical Analysis?

Whether you're interested in using StickForStats for your research, contributing to the project, or discussing potential collaborations, I'd love to hear from you.

Email Me GitHub