Part of the Country Resilience Score feature. See full plan: docs/plans/2026-03-29-feat-country-resilience-score-plan.md

Scope

Port 6 pure functions from qadr110's nrc-statistics.ts into server/_shared/resilience-stats.ts. No Redis reads, no side effects.

File: server/_shared/resilience-stats.ts

Functions:

• cronbachAlpha(items: number[][]): number — composite reliability check, flags low-confidence scores
• detectTrend(values: number[]): 'rising' | 'stable' | 'falling' — linear regression slope
• detectChangepoints(values: number[], threshold?): number[] — CUSUM structural break detection
• minMaxNormalize(values: number[]): number[] — dimension normalization utility
• exponentialSmoothing(values: number[], alpha?): number[] — smoothing noisy real-time inputs
• nrcForecast(history, horizonDays, alpha?): { values, confidenceIntervals, probabilityUp, probabilityDown } — hybrid ETS + linear trend

Do NOT port: deaEfficiency, controlChart, laggedCorrelation.

Tests: tests/resilience-stats.test.mts

• cronbachAlpha: known matrix, single-row, all-same values
• detectTrend: monotone up/down/flat, fewer than 3 values
• detectChangepoints: bimodal series, constant series, short series
• minMaxNormalize: boundary values (all same, empty, negatives)
• nrcForecast: output length = horizonDays, CI lower ≤ upper, values in 0–100, probabilities sum to 1

Done when

All tests pass, typecheck:api clean.