MomentBalance 제작자: Jason

MomentBalance is a professional research tool that creates statistically balanced experimental groups from CSV measurement data using high-performance parallel optimization algorithms.

Core Functionality

• CSV Data Import: Upload files with subject IDs and measurement values (volume in mm³, weights, biomarkers, etc.)
• Dual Optimization: Balances both group means AND standard deviations simultaneously—crucial for reducing confounding variables
• Parallel WASM Processing: WebAssembly-compiled Rust algorithm with Web Worker pool for true multi-core parallelism
• Simulated Annealing Algorithm: Temperature-controlled probabilistic optimization, calibrated per dataset and targeting a fixed wall-clock time budget rather than a fixed iteration count
• Flexible Group Configuration: Configure any number of groups from 2 up to your specimen count, with automatic size distribution
• Quick Preview: A faster, rougher search for iterating on your group count before committing to the full run

Advanced Features

• High-Performance Computing: 20-25× faster than pure JavaScript through compiled WASM and parallel execution
• Multi-Core Utilization: Automatically scales to available CPU cores for optimal performance
• Multilingual Interface: Full support for Canadian English, Canadian French, German, Italian, and Japanese with automatic browser locale detection
• Statistical Validation: Displays a citable ANOVA p-value, optimizer fitness scores, and per-variable variance metrics
• Progress Tracking: Real-time progress feedback with optional completion notification
• Semantic UI: Clean, accessible interface with automatic dark/light theme using Pico CSS Classless
• Export-Ready Results: Detailed group assignments with individual measurements and statistical summaries in CSV format

Research Applications

• Animal Studies: Balance tumour volumes, body weights, biomarker levels
• Clinical Trials: Create matched patient groups based on baseline measurements
• Agricultural Research: Balance soil samples, plant measurements, treatment plots
• Manufacturing QC: Balance product batches, material properties, testing samples
• Educational Research: Create equivalent student groups for intervention studies

Technical Specifications

• Architecture: Rust/WASM computation engine with JavaScript UI layer
• Parallelization: Worker pool distributing independent optimization runs across CPU cores
• Optimization Target: Minimizes variance of group averages + variance of group standard deviations, normalized per variable so differently-scaled measurements are weighted equally
• Algorithm: Greedy initialization followed by temperature-controlled simulated annealing, with multiple independent trials run in parallel and the best result kept
• Processing Budget: Targets a fixed wall-clock time (about one second of parallel processing) rather than a fixed iteration count or user-facing quality setting; iteration depth is computed automatically from specimen count and available CPU cores
• Package Size: approximately 148 KB total—includes the WASM engine, UI framework, 5 language files, and all dependencies
• Data Format: Simple CSV (comma-separated, no quoted fields)
• Localization: Locale-aware number formatting (decimal separators, thousands grouping, scientific notation)

Why MomentBalance?

Traditional random assignment often creates unbalanced groups, introducing confounding variables that reduce statistical power. MomentBalance ensures your experimental groups start with equivalent statistical properties, improving the validity of your analyses and reducing the number of subjects needed to detect effects.

The dual optimization of both means and standard deviations is particularly important when baseline variability differs between potential group assignments—a common issue in biological research where individual variation is high.

The parallel WASM architecture enables processing larger datasets than JavaScript-based tools, completing typical analyses in about a second of wall-clock time.

Perfect for researchers, statisticians, and quality control professionals requiring rigorous experimental design with reproducible group assignments.