@hugov/correl-range2/src/_random-number.js

monte carlo simulation for correlated variables expressed as ranges

export default class RandomNumber {
	/**
	 * @param {([s: number]) => number} fz - Z random number gererator with optional [0-1] random seed
	 */
	constructor(fz) {
		this._fz = fz
		this._ks = [] // risk index of weights ws
		this._ws = [] // weights for risk indices ks
		this.value = NaN
	}
	valueOf() {
		return this.value
	}
	/**
	 * @param {Array<number>} zs - Z random iid numbers
	 */
	update(zs) {
		let v = 0
		for (var i=0; i<this._ks.length; ++i) v += this._ws[i] * zs[this._ks[i]]
		this.value = this._fz(v)
		return this
	}
	/**
	 * TODO - custom language in tag template: L`1 2 economy 3%` vs L(1,2,'economy',.03)
	 * @param {Array<string>} risks - random iid names|indices
	 * @param {Object} factors - name-weight risks
	 */
	_link(risks, factors) {
		const ks = this._ks,
					ws = this._ws
		let Δ = 1,
				i = 0
		Object.keys(factors).forEach(risk => {
			ks.push( riskIndex( risks, risk ) )
			const w = factors[risk]
			Δ -= (ws[ws.length] = w)**2
			if (Δ < -Number.EPSILON) throw Error('sum of squared weights > 1')
		})
		// only bother is there is some weight to be assigned
		if (Δ > Number.EPSILON) {
			ks.push( risks.push('self') - 1 )
			ws.push( Math.sqrt( Δ ) )
		}
		return this
	}
}
function riskIndex(risks, riskName) {
	let idx = risks.indexOf(riskName)
	return idx !== -1 ? idx : risks.push(riskName ?? '') - 1
}