Least Squares Solution Calculator

Example Data Table

This example fits a simple straight line using one intercept column and one predictor column.

Formula Used

The least squares method finds a vector that minimizes the squared error between actual values and predicted values.

min ||Ax - b||²

For ordinary least squares, the normal equation is:

x = (AᵀA)⁻¹Aᵀb

When ridge lambda is greater than zero, the calculator uses:

x = (AᵀA + λI)⁻¹Aᵀb

The residual for each observation is:

residual = actual b - fitted value

How to Use This Calculator

1. Enter matrix A with each row on a new line.
2. Enter vector b with matching row count.
3. Use commas, spaces, or semicolons as separators.
4. Enable the intercept option when your matrix lacks a constant column.
5. Set ridge lambda above zero for unstable or nearly singular systems.
6. Press the calculate button.
7. Review coefficients, fitted values, residuals, and error metrics.
8. Download the result as CSV or PDF.

Least Squares Solution Guide

What This Calculator Solves

A least squares solution is useful when a system has more equations than unknowns. Such systems often have no exact answer. The calculator finds the best approximate vector. It makes the total squared residual as small as possible. This is common in statistics, regression modeling, curve fitting, calibration, forecasting, and data analysis.

Why Residuals Matter

Residuals show the gap between observed values and fitted values. A small residual means the model fits that observation well. A large residual may reveal an outlier, weak model form, missing predictor, or measurement issue. The residual chart helps you inspect these errors quickly.

Understanding the Coefficients

Each coefficient explains the weight assigned to a column of matrix A. In regression, a coefficient shows how much the fitted value changes when that predictor changes. The intercept option adds a constant term. This is helpful when your data needs a baseline value instead of forcing the fitted line through zero.

Using Ridge Stabilization

Some matrices create unstable normal equations. This happens when columns are highly related or nearly dependent. Ridge lambda adds a small value to the diagonal of A transpose A. This can improve numerical stability. Use a small lambda first. Large lambda values may shrink coefficients too strongly.

Reading the Metrics

SSE is the total squared error. MSE is the average adjusted squared error. RMSE is easier to read because it uses the same unit as b. R squared describes how much variation is explained by the fitted model. These metrics should be read together, not alone.

FAQs