Combining measurements of different precision by weighting each by the inverse square of its uncertainty.The weighted mean is a method for combining several measurements of the same quantity when those measurements have different precisions. Instead of taking a simple arithmetic average, each measurement is multiplied by a weight that is inversely proportional to its variance (the square of its uncertainty). This gives more influence to measurements with smaller uncertainties, which are more reliable. The result is the best linear unbiased estimate of the true value, assuming the measurements are independent and their errors are Gaussian.
The concept of weighting by inverse variance dates back to the early 19th century, with contributions from Gauss and Legendre in the context of least squares. The explicit formula for the weighted mean is a standard result in error theory.
For each measurement x_i with uncertainty σ_i, the weight w_i is set to 1/σ_i². The weighted mean is then the sum of (w_i * x_i) divided by the sum of w_i. The uncertainty of the weighted mean is the square root of the reciprocal of the sum of the weights. This formula emerges naturally from maximum likelihood estimation when the errors are Gaussian, and it is also the solution that minimizes the chi-squared for a constant model.
Used in particle physics to combine measurements of particle masses from different experiments, in metrology to define international standards, and in astronomy to combine observations from different telescopes.
If all uncertainties are equal, the weighted mean reduces to the ordinary arithmetic mean, so the weighted mean is a strict generalization.







