statistical-analysislisted

# Statistical Analysis Service v4 ## Core Philosophy: Diagnosis Before Analysis v3 was "do whatever the user asks"; v4 is "diagnose the data and assumptions first, then decide what to do." ``` v3: User request → Select path → Execute v4: User request → Data profile → Assumption checks → Intelligent method selection → Execute → Triplet output (table + figure + paragraph) ``` --- ## Workflow Overview ``` User request | v +-------------------------------------+ | Step 0: Data Profile (required for | | all paths) | | - Sample size, variable types, | | missing patterns | | - Distribution characteristics, | | outlier detection | | - Completed within 30 seconds, | | no confirmation needed | +-------------------------------------+ | v Assess complexity -> Select path | |-- Quick path -> Assumption self-check -> Execute -> Triplet output |-- Light path -> Assumption self-check -> Confirm variables -> Execute -> Triplet output +-- Full path -> Four stages (incl. assumption checks) -> Triplet output ``` --- ## Step 0: Data Profile **Must be executed before all analyses**. Output format: ```markdown ## Data Profile | Metric | Value | |--------|-------| | Sample size | N = 3248 | | Variables | 94 (continuous: 60, categorical: 34) | | Missing rate | Overall 2.3%, highest: Variable X (15.2%) | | Outliers | Variable Y has 12 (> 3SD) | ### Key Variable