Skip to content

Latest commit

 

History

History
96 lines (81 loc) · 7.87 KB

File metadata and controls

96 lines (81 loc) · 7.87 KB

Probability & Statistics — Formulas

Quick reference of the formulas used across the weekly notes, with brief explanations and symbol keys. See weekly notes for full context and examples.

Week 01 — Descriptive stats

  • Sample mean: $\bar{x} = \frac{1}{n}\sum_{j=1}^n x_j$. Average of the sample values.
  • Median (odd $n$): $m = y_{(n+1)/2}$; (even $n$): $m = \frac{y_{n/2}+y_{n/2+1}}{2}$. Middle ordered value(s).
  • Sample variance / SD: $s^2 = \frac{1}{n-1}\sum_{j=1}^n (x_j-\bar{x})^2$, $s=\sqrt{s^2}$. Uses $n-1$ to estimate population variance from a sample.
  • Mean absolute deviation: $\text{m.a.d.} = \frac{1}{n}\sum_{j=1}^n |x_j-\bar{x}|$. Average absolute distance from the mean.
  • Empirical CDF: $F_{\text{emp}}(x)=\frac{1}{n}#{j: x_j\le x}$. Fraction of data at or below $x$.
    • Symbols: $x_j$ data points, $n$ sample size, $y_j$ ordered values, $#{\cdot}$ count.

Week 02 — Probability basics

  • Kolmogorov axioms: $P(\varnothing)=0$, $P(\Omega)=1$, $P(\bigcup A_j)=\sum P(A_j)$ for disjoint $A_j$.
  • Combinatorial probability (finite uniform): $P(E)=\frac{#E}{n}$.
  • Inclusion–exclusion (two events): $P(A\cup B)=P(A)+P(B)-P(A\cap B)$.
  • Independence: $P(A\cap B)=P(A)P(B)$.
  • Conditional probability: $P(A|B)=\frac{P(A\cap B)}{P(B)}$.
  • Total probability / Bayes: $P(A)=\sum_j P(A|B_j)P(B_j)$; $P(B_j|A)=\frac{P(A|B_j)P(B_j)}{\sum_k P(A|B_k)P(B_k)}$.
    • Symbols: $A,B,B_j$ events; $\Omega$ sample space; $#E$ count; $n$ outcomes.

Week 03 — Random variables (discrete)

  • CDF: $F_X(x)=P(X\le x)$ (step function for discrete $X$).
  • PMF: $p_X(x)=P(X=x)$ with $\sum_x p_X(x)=1$.
  • PMF/CDF link (discrete): jump at $x_j$ equals $p_X(x_j)=F_X(x_j)-\lim_{x\to x_j^-}F_X(x)$.
    • Symbols: $X$ random variable; $p_X$ pmf; $F_X$ cdf; $x_j$ support points.

Week 04 — Expectation, variance, covariance

  • Expectation (discrete): $E\xi=\sum_x x,p_\xi(x)$; of a function: $E g(\xi)=\sum_x g(x)p_\xi(x)$.
  • Variance: $\text{Var}(\xi)=E(\xi-E\xi)^2=E\xi^2-(E\xi)^2$.
  • Linearity: $E\left(\sum_j \xi_j\right)=\sum_j E\xi_j$; $\text{Var}(c\xi)=c^2\text{Var}(\xi)$; for independent $\xi_j$, $\text{Var}(\sum_j \xi_j)=\sum_j \text{Var}(\xi_j)$.
  • Covariance / correlation: $\text{Cov}(\xi_1,\xi_2)=E\xi_1\xi_2-E\xi_1E\xi_2$; $\rho=\frac{\text{Cov}(\xi_1,\xi_2)}{\sqrt{\text{Var}(\xi_1)\text{Var}(\xi_2)}}$.
  • Markov: $P(\xi\ge x)\le \frac{E\xi}{x}$ for $\xi\ge0$; Chebyshev: $P(|\xi-E\xi|\ge x)\le \frac{\text{Var}(\xi)}{x^2}$.
    • Symbols: $\xi,\xi_j$ random variables; $p_\xi$ pmf; $g$ measurable function; $c$ constant; $x>0$ bound.

Week 05 — Continuous r.v.s, pdf/cdf, normal

  • PDF/CDF link: $F_\xi(x)=\int_{-\infty}^x p_\xi(t),dt$; if $F_\xi$ differentiable, $p_\xi(x)=F'_\xi(x)$.
  • Uniform $(a,b)$: $p_\xi(x)=\frac{1}{b-a}I(a<x<b)$; $F_\xi(x)=0$ ($x\le a$), $(x-a)/(b-a)$ ($a<x<b$), $1$ ($x\ge b$); $E\xi=\frac{a+b}{2}$, $\text{Var}(\xi)=\frac{(b-a)^2}{12}$.
  • Normal $(\mu,\sigma^2)$ pdf: $p_\xi(x)=\frac{1}{\sqrt{2\pi}\sigma}\exp\left(-\frac{(x-\mu)^2}{2\sigma^2}\right)$; standardization $\eta=\frac{\xi-\mu}{\sigma}\sim N(0,1)$.
    • Symbols: $p_\xi$ pdf; $F_\xi$ cdf; $I(\cdot)$ indicator; $\mu,\sigma$ mean/SD; $a,b$ interval endpoints.

Week 06 — Hypergeometric, exponential, Cauchy, transforms

  • Hypergeometric pmf: $P(\xi=x)=\frac{\binom{b}{x}\binom{N-b}{n-x}}{\binom{N}{n}}$; $E\xi=\frac{bn}{N}$; $\text{Var}(\xi)=\frac{bn(N-b)(N-n)}{N^2(N-1)}$.
  • Exponential pdf: $p_\xi(x)=\lambda e^{-\lambda x}I(x\ge0)$; $E\xi=1/\lambda$; $\text{Var}(\xi)=1/\lambda^2$; memoryless $P(\xi>t+s|\xi>t)=P(\xi>s)$.
  • Cauchy pdf: $p_\xi(x)=\frac{1}{\pi(1+x^2)}$ (no finite moments).
  • Expectation via density: $E\xi=\int x p_\xi(x),dx$; $E g(\xi)=\int g(x)p_\xi(x),dx$.
  • Density of $f(\xi)$ (invertible $f$): $p_\eta(y)=p_\xi(f^{-1}(y))\left|\frac{d}{dy}f^{-1}(y)\right|I(y\in \text{image})$.
    • Symbols: $N$ population size, $b$ good items, $n$ draws; $\lambda$ rate; $f$ transformation; $\eta=f(\xi)$.

Week 07 — Poisson, joint pdfs, conditionals, transforms

  • Poisson pmf: $P(\xi=k)=\frac{\lambda^k e^{-\lambda}}{k!}$ for $k\in\mathbb{N}$.
  • Joint pdf to marginal: $p_{\xi_1}(x_1)=\int p_{\xi}(x_1,x_2),dx_2$ (similarly for other components).
  • Conditional pdf (continuous): $p_{\xi_2|\xi_1}(x_2|x_1)=\frac{p_{\xi}(x_1,x_2)}{p_{\xi_1}(x_1)}$.
  • Transformation with Jacobian (invertible map $T$): $p_\eta(y)=p_\xi(T^{-1}(y))\left|\det J_{T^{-1}}(y)\right|$.
    • Symbols: $\lambda$ rate; $p_\xi$ joint pdf; $J_{T^{-1}}$ Jacobian of inverse map; $\eta=T(\xi)$.

Week 09 — Multivariate normal, chi-squared, sampling

  • Multivariate normal pdf: $p_\xi(x)=\frac{1}{(2\pi)^{d/2}\sqrt{\det K}}\exp\left(-\frac{1}{2}(x-\mu)^T K^{-1}(x-\mu)\right)$.
  • Chi-squared (sum of squares of $d$ iid $N(0,1)$): cdf/pdfs as special case (noted in lecture).
  • Inversion sampling: $\xi=F^{-1}(U)$ with $U\sim \text{Uni}(0,1)$ gives cdf $F$.
  • Exponential sampling via inversion: $X=-\frac{1}{\lambda}\ln U$ has $F(x)=1-e^{-\lambda x}$.
    • Symbols: $\mu$ mean vector, $K$ covariance matrix, $d$ dimension; $U$ uniform(0,1).

Week 10 — Couplings, transport (selected formulas)

  • Standardization of sums (CLT-style): $\frac{S_n-n\mu}{\sigma\sqrt{n}}$ used for normal approximations.
  • Transport cost intuition (no specific closed form given in notes; see week10 text for details).
    • Symbols: $S_n$ sum, $\mu$ mean, $\sigma$ SD.

Week 11 — WLLN, Bernstein, CLT

  • Bernstein bound (Bernoulli sample mean): $P\left(\left|\frac{S_n}{n}-p\right|\ge\varepsilon\right)\le 2e^{-n\varepsilon^2/4}$.
  • Chebyshev for Bernoulli: $P\left(\left|\frac{S_n}{n}-p\right|\ge\varepsilon\right)\le \frac{p(1-p)}{n\varepsilon^2}$.
  • CLT normalization: $\frac{S_n-n\mu}{\sigma\sqrt{n}}\xrightarrow{d}N(0,1)$; sample mean version $\frac{\bar{\xi}_n-\mu}{\sigma/\sqrt{n}}\xrightarrow{d}N(0,1)$.
    • Symbols: $S_n$ sum of iid with mean $\mu$, variance $\sigma^2$; $p$ Bernoulli parameter; $\varepsilon>0$ tolerance.

Week 12 — Estimation and confidence intervals

  • Shifted exponential-type CI (minimum-based): for pdf $e^{-(x-\theta)}I(x\ge\theta)$, with sample minimum $m$, choose $c_\alpha$ s.t. $P(m-\theta\ge c_\alpha)=\alpha$; interval $(m-c_\alpha,,m)$ has level $1-\alpha$.
  • Normal mean, variance known: $\bar{\xi}n \pm z{1-\alpha/2}\frac{\sigma}{\sqrt{n}}$ is a $(1-\alpha)$ CI for $\theta$.
  • Normal variance, mean known: based on $\chi^2$ (see week12 text for exact quantiles).
    • Symbols: $\bar{\xi}n$ sample mean; $\sigma$ known SD; $z{1-\alpha/2}$ standard normal quantile; $m$ sample minimum; $\alpha$ significance.

Week 13 — Risk, UMVU, median asymptotics

  • Risk (squared error): $R(\hat{\theta},\theta)=E_\theta[(\hat{\theta}-\theta)^2]$.
  • Sample median definition: for ordered $\xi_{(1)}\le\dots\le\xi_{(n)}$, median is $\xi_{(j+1)}$ if $n=2j+1$; $(\xi_{(j)}+\xi_{(j+1)})/2$ if $n=2j$.
  • Asymptotic normality (informal): $\sqrt{n}(\text{median}-\theta)\xrightarrow{d}N\left(0,\frac{1}{4p_\xi(\theta)^2}\right)$ when density at median $\theta$ is positive.
    • Symbols: $\hat{\theta}$ estimator; $\theta$ true parameter; $p_\xi$ density at median.

Week 14 — Method of moments, MLE, hypothesis testing

  • Sample moments: $M_k=\frac{1}{n}\sum_{j=1}^n \xi_j^k$; method-of-moments solves $\alpha_k(\theta)=M_k$.
  • Likelihood (iid): $L(\theta)=\prod_{j=1}^n p_\xi(x_j;\theta)$; log-likelihood $\ell(\theta)=\sum_{j=1}^n \ln p_\xi(x_j;\theta)$.
  • One-sided normal test (variance known): reject $H_0:\theta=\theta_0$ if $\bar{\xi}n\ge \theta_0+z{1-\alpha}\frac{\sigma}{\sqrt{n}}$.
    • Symbols: $\xi_j$ sample; $\alpha_k$ theoretical moments; $\theta$ parameter(s); $p_\xi$ model density/pmf; $\sigma$ known SD; $z_{1-\alpha}$ quantile.

Week 15 — Hypothesis testing II

  • Power function (conceptual): $\beta(\theta)=P_\theta(\text{reject }H_0)$; aim for small Type I, large power on $H_1$.
  • Unbiased/consistent tests: see week15 notes for specific constructions; formulas follow the same critical-value pattern as week14.
    • Symbols: $\beta(\theta)$ power; $\alpha$ significance; $H_0,H_1$ hypotheses.