SPB Git

Repository Public

distribution.tsx 7.0 KB 138 lines tsx

1import type { Locale } from "@/schema";2import { linspace, normalCdf, normalPdf, studentTCdf, studentTPdf } from "../core";3import { option, optionError, param, type ParamDef, type SimConfig, type SimDef } from "../types";4import { fmt } from "../ui/format";5import { Plot } from "../ui/Plot";6import { Readouts } from "../ui/Readouts";78export const FAMILIES = ["normal", "student", "lognormal", "exponential", "uniform"] as const;9export type Family = (typeof FAMILIES)[number];1011interface FamilyDef {12  domain: [number, number];13  params: ParamDef[];14  pdf(x: number, p: Readonly<Record<string, number>>): number;15  cdf(x: number, p: Readonly<Record<string, number>>): number;16  mean(p: Readonly<Record<string, number>>): number | null;17  variance(p: Readonly<Record<string, number>>): number | null;18}19const ab = (lo: number, hi: number, a: number, b: number): ParamDef[] => [20  param("a", "Borne a", "Bound a", lo, hi, 0.1, a), param("b", "Borne b", "Bound b", lo, hi, 0.1, b),21];22export const FAMILY_DEFS: Record<Family, FamilyDef> = {23  normal: {24    domain: [-8, 8],25    params: [param("mu", "Moyenne μ", "Mean μ", -3, 3, 0.1, 0), param("sigma", "Écart-type σ", "Standard deviation σ", 0.3, 3, 0.1, 1), ...ab(-8, 8, -1, 1)],26    pdf: (x, p) => normalPdf(x, p.mu, p.sigma), cdf: (x, p) => normalCdf(x, p.mu, p.sigma),27    mean: (p) => p.mu, variance: (p) => p.sigma * p.sigma,28  },29  student: {30    domain: [-8, 8],31    params: [param("df", "Degrés de liberté ν", "Degrees of freedom ν", 1, 30, 1, 4), ...ab(-8, 8, -2, 2)],32    pdf: (x, p) => studentTPdf(x, p.df), cdf: (x, p) => studentTCdf(x, p.df),33    mean: (p) => (p.df > 1 ? 0 : null), variance: (p) => (p.df > 2 ? p.df / (p.df - 2) : null),34  },35  lognormal: {36    domain: [0, 10],37    params: [param("mu", "μ de ln X", "μ of ln X", -1, 1.5, 0.1, 0.5), param("sigma", "σ de ln X", "σ of ln X", 0.1, 1.5, 0.05, 0.5), ...ab(0, 10, 1, 3)],38    pdf: (x, p) => (x > 0 ? normalPdf(Math.log(x), p.mu, p.sigma) / x : 0),39    cdf: (x, p) => (x > 0 ? normalCdf(Math.log(x), p.mu, p.sigma) : 0),40    mean: (p) => Math.exp(p.mu + (p.sigma * p.sigma) / 2),41    variance: (p) => (Math.exp(p.sigma * p.sigma) - 1) * Math.exp(2 * p.mu + p.sigma * p.sigma),42  },43  exponential: {44    domain: [0, 10],45    params: [param("lambda", "Taux λ", "Rate λ", 0.2, 5, 0.1, 1), ...ab(0, 10, 0, 1)],46    pdf: (x, p) => (x >= 0 ? p.lambda * Math.exp(-p.lambda * x) : 0), cdf: (x, p) => (x > 0 ? 1 - Math.exp(-p.lambda * x) : 0),47    mean: (p) => 1 / p.lambda, variance: (p) => 1 / (p.lambda * p.lambda),48  },49  uniform: {50    domain: [-6, 9],51    params: [param("low", "Minimum", "Minimum", -5, 3, 0.1, 0), param("width", "Largeur", "Width", 0.5, 5, 0.1, 2), ...ab(-6, 9, 0.5, 1.5)],52    pdf: (x, p) => (x >= p.low && x <= p.low + p.width ? 1 / p.width : 0),53    cdf: (x, p) => Math.min(1, Math.max(0, (x - p.low) / p.width)),54    mean: (p) => p.low + p.width / 2, variance: (p) => (p.width * p.width) / 12,55  },56};5758export interface DistributionModel {59  family: Family;60  domain: [number, number];61  curve: { x: number; y: number }[];62  /** Interval after ordering the two bounds. */63  a: number; b: number;64  /** Density over [a, b], for the shaded area. */65  shade: { x: number; lo: number; hi: number }[];66  /** P(a ≤ X ≤ b) = F(b) − F(a), from the closed-form distribution function. */67  probability: number;68  mean: number | null; sd: number | null;69}7071export function computeDistribution(p: Readonly<Record<string, number>>, config: SimConfig): DistributionModel {72  const family = option(config, "family", FAMILIES);73  const F = FAMILY_DEFS[family];74  const a = Math.min(p.a, p.b), b = Math.max(p.a, p.b);75  const grid = linspace(F.domain[0], F.domain[1], 321);76  let curve = grid.map((x) => ({ x, y: F.pdf(x, p) }));77  if (family === "uniform") {78    const h = 1 / p.width, hi = p.low + p.width;79    curve = [{ x: F.domain[0], y: 0 }, { x: p.low, y: 0 }, { x: p.low, y: h }, { x: hi, y: h }, { x: hi, y: 0 }, { x: F.domain[1], y: 0 }];80  }81  const inside = [a, ...grid.filter((x) => x > a && x < b), b];82  const variance = F.variance(p);83  return {84    family, domain: F.domain, curve, a, b,85    shade: inside.map((x) => ({ x, lo: 0, hi: F.pdf(x, p) })),86    probability: Math.max(0, F.cdf(b, p) - F.cdf(a, p)),87    mean: F.mean(p), sd: variance === null ? null : Math.sqrt(variance),88  };89}9091const NAMES: Record<Family, { fr: string; en: string }> = {92  normal: { fr: "Loi normale", en: "Normal distribution" }, student: { fr: "Loi de Student", en: "Student-t distribution" },93  lognormal: { fr: "Loi log-normale", en: "Log-normal distribution" }, exponential: { fr: "Loi exponentielle", en: "Exponential distribution" },94  uniform: { fr: "Loi uniforme", en: "Uniform distribution" },95};96const TXT = {97  fr: { density: "densité f(x)", area: "P(a ≤ X ≤ b)", prob: "P(a ≤ X ≤ b)", mean: "Espérance", sd: "Écart-type", family: "Loi", none: "non définie" },98  en: { density: "density f(x)", area: "P(a ≤ X ≤ b)", prob: "P(a ≤ X ≤ b)", mean: "Mean", sd: "Standard deviation", family: "Distribution", none: "undefined" },99} as const;100101function DistributionView({ model: m, locale, idPrefix }: { model: DistributionModel; locale: Locale; idPrefix: string }) {102  const L = TXT[locale];103  return (104    <>105      <Plot locale={locale} idPrefix={idPrefix} spec={{106        title: NAMES[m.family][locale], xLabel: "x", yLabel: L.density, xDomain: m.domain, yFromZero: true,107        layers: [108          { type: "band", points: m.shade, tone: "blue", name: L.area, opacity: 0.28 },109          { type: "line", points: m.curve, tone: "blue", name: L.density },110          { type: "vline", x: m.a, tone: "neutral", label: "a" }, { type: "vline", x: m.b, tone: "neutral", label: "b" },111        ],112      }} />113      <Readouts items={[114        { label: L.family, value: NAMES[m.family][locale] },115        { label: L.prob, value: fmt(locale, m.probability, 4), tone: "blue" },116        { label: L.mean, value: m.mean === null ? L.none : fmt(locale, m.mean, 3) },117        { label: L.sd, value: m.sd === null ? L.none : fmt(locale, m.sd, 3) },118      ]} />119    </>120  );121}122123export const distribution: SimDef<DistributionModel> = {124  kind: "distribution",125  title: { fr: "Loi de probabilité et probabilité d'un intervalle", en: "Probability distribution and interval probability" },126  description: {127    fr: "Densité de la loi choisie et probabilité $P(a \\le X \\le b) = F(b) - F(a)$, calculée avec la fonction de répartition exacte (aucun tirage). L'espérance et l'écart-type sont ceux de la loi ; pour Student ils n'existent que si $\\nu > 1$ et $\\nu > 2$.",128    en: "Density of the chosen distribution and the probability $P(a \\le X \\le b) = F(b) - F(a)$, computed from the exact distribution function (no random draw). Mean and standard deviation are those of the distribution; for Student-t they exist only when $\\nu > 1$ and $\\nu > 2$.",129  },130  params: FAMILY_DEFS.normal.params,131  stochastic: false,132  configKeys: ["family"],133  paramsFor: (c) => FAMILY_DEFS[option(c, "family", FAMILIES)].params,134  validateConfig: (c) => optionError(c, "family", FAMILIES),135  compute: (p, _seed, c) => computeDistribution(p, c),136  View: DistributionView,137};138
Files main
Menu

Theme