A professional-grade predictive analytics platform and statistical modeling suite. Perform high-precision hypothesis testing and machine learning with local AI insights powered by Ollama—ensuring 100% data privacy and secure research workflows.
The Statistics Toolkit is designed for researchers, data scientists, and students who require rigorous statistical validation without sacrificing privacy.
| Category | Description | Primary Analytical Tools |
|---|---|---|
| Predictive Suite | Advanced Machine Learning for forecasting. | Linear/Logistic Regression, Feature Importance, Cross-Validation. |
| Inferential Logic | Scientific Hypothesis Testing & Validation. | Independent T-Test, One-Way ANOVA, Assumption Auditing. |
| AI Consultant | Private AI-driven results interpretation. | "Which Test?" Wizard, Automated APA 7th Results Writer, AI Dataset Profiler & Clean. |
| Descriptive Data | Comprehensive summary statistics. | Mean, Median, Skewness, Kurtosis, Outlier Detection. |
| Data Persistence | Secure result management & portability. | CSV Research Export, SQLite Session History. |
- 🛡️ Automated Assumption Audit: Never run a biased test again. Every inference tool automatically validates Normality (Shapiro-Wilk) and Homogeneity (Levene's) before presenting results.
- 🤖 Local & Cloud AI Insights: Supports a dual-channel hybrid routing engine. Seamlessly connects to direct cloud endpoints or falls back to a local Ollama server. Provides plain-English interpretations of complex stats.
- 📄 Automated APA 7th Results Writer: Instantly drafts flawless, publication-ready "Results and Analysis" sections adhering strictly to the APA 7th Edition manual guidelines, formatting standard italicized notations (t, F, p, d, df) with precise statistical figures.
- 📊 AI Dataset Profiling & Clean: Probes variable structures, identifies skewness and distribution anomalies, runs IQR outlier detection, and suggests target-specific cleaning protocols inside an immersive, glassmorphic viewport.
- 🧠 Local Hybrid RAG Engine: Features a Retrieval-Augmented Generation (RAG) pipeline. AI recommendations and results interpretation are grounded in a local statistical knowledge base (e.g., APA 7th standards, test decision matrixes).
- Semantic Embedding Match: Matches search queries with local text chunks using local embedding vectors.
- Zero-Dependency TF-IDF Fallback: Automatically transitions to an optimized, pure-Python search index if the local embedding API is offline.
- 📈 Model Audit Dashboard: Visualize Feature Impact and Error Metrics (MAE, RMSE, R²) in real-time to understand what drives your predictions.
- ♿ Inclusive Design: Fully compliant with WCAG 2.2 AA standards, featuring aria-live regions, high-contrast themes, and full keyboard navigation support.
- 📥 Professional Data Export: Export your entire calculation history as a research-ready CSV file for use in academic manuscripts or external BI tools.
- Backend: Python 3.12+ (Flask)
- Analytics: Scikit-Learn, SciPy, NumPy, Statsmodels
- AI Engine: Ollama (Local Model Integration)
- Frontend: Vanilla JS, Chart.js
- Accessibility: WCAG 2.2 AA Standardized
- Install Ollama: Visit ollama.com and install the local server.
- Pull the Model: Run
ollama pull phi3.5:3.8b. - Setup Environment:
# 1. Install dependencies pip install -r requirements.txt # 2. Run the app flask run
We are continuously evolving the toolkit to meet the needs of modern researchers. Upcoming features include:
- 📊 Advanced Visual Diagnostics: Integration of Violin plots, Q-Q plots, and Residual dashboards for deeper model validation.
- ⚖️ Bayesian Module: Moving beyond frequentist p-values with Bayesian credible intervals and posterior distribution modeling.
- 📉 Time-Series Forecasting: Implementation of ARIMA and Prophet models for sequential data analysis and trend prediction.
- 🧠 Deep Learning Integration: A neural network wizard for building simple Keras/PyTorch models with automated hyperparameter tuning.
- 🔐 Encrypted Cloud Sync: Optional, end-to-end encrypted synchronization for multi-device research collaboration.
Developed for researchers who value privacy, precision, and inclusive design.