StatMath

Notes on statistics & applied mathematics

A personal, slowly growing collection of worked notes, derivations and references — from descriptive statistics and probability to regression and numerical methods. Written to be re-read, not to impress.

Topics

Descriptive statistics

Summaries, moments, quantiles, robust measures of spread, and when the mean lies.

Probability

Distributions, expectation, the law of large numbers and the central limit theorem.

Estimation & inference

Maximum likelihood, confidence intervals, hypothesis tests and their pitfalls.

Regression

Linear and logistic models, diagnostics, regularization and interpretation.

Numerical methods

Root finding, quadrature, linear algebra and stability of algorithms.

Bayesian methods

Priors, posteriors, conjugacy and a first look at MCMC sampling.

Recent notes

14 Oct 2026 · probability

The central limit theorem, illustrated

Why sums of independent variables drift toward a normal shape — regardless of the distribution you started with. A short derivation and a simulation.

Zn = ( X̄n − μ ) / ( σ / √n )  ⟶  N(0, 1)  as n → ∞
Histogram of 10 000 sample means (n = 30) with the fitted normal density.
28 Sep 2026 · inference

What a 95% confidence interval does and does not say

The interval is random; the parameter is not. A note on the interpretation people get wrong most often, with a coverage simulation.

x̄ ± tα/2,  n−1 · s / √n
2 Aug 2026 · bayesian

Bayes' rule as bookkeeping

Updating a Beta prior with Bernoulli data stays Beta — a clean example of conjugacy that makes the mechanics obvious.

p(θ | x) ∝ p(x | θ) · p(θ)
11 Jun 2026 · regression

Reading residual plots

Most of what a linear model is hiding shows up in its residuals. A checklist of the four patterns worth learning to recognize at a glance.

Resources & references

About

StatMath is a personal notebook kept public mostly so the formatting stays honest. The notes are written while working through problems and revised when they turn out to be wrong. personalnon-commercialupdated occasionally

Corrections and suggestions are welcome by email.