Sandhya Indurkar

Math foundations

Matrices and Linear Systems: Regression Is Ax = b

Matrix A times vector x equals vector b

The idea

A matrix is a compact way to write many linear equations at once. Each row can be one data point constraint; each column is one unknown coefficient. Fitting a line through points is solving for intercept and slope that best satisfy all rows.

Linear systems answer: What values of x make Ax as close as possible to b across all rows?

Example: Ax = b in 2×2 form

Two equations, two unknowns. Regression normal equations use the same structure with more rows.

Two weeks, two levers: solve for intercept and ad slope from normal equations.

Matrix A

[4.0, 10.0]

[10.0, 30.0]

Vector b

[20.0]

[58.0]

intercept a

1.0

slope b

1.6

Solution: intercept a = 2.5, slope b = 1.0. Regression with two unknowns (intercept + slope) is the same pattern: matrix A from data sums, vector b from targets.

The math

System form

Ax = b

A holds coefficients, x holds unknowns, b holds targets or totals from data.

Regression notation

Xβ ≈ y

Design matrix X (rows = weeks, columns = intercept + drivers), coefficient vector β, outcome vector y. Same structure as Ax = b.

Least squares solution

normal equations: XᵀX β = Xᵀy

When rows outnumber unknowns, you solve this system instead of exact equality. That is what your regression code implements under the hood.

A simple application

When someone asks what regression did, say it found β that solves a linear system built from your data table. The linear models post draws the line; this post names the matrix behind it.