Least absolute deviations fitting (Bošković)
Ajuste por mínimas desviaciones absolutas (Bošković)
- Date
- 1755-1757 · confidence medium
- Place
- Rome, Italy
- Credited to
- Ruđer Bošković
- Field
- Knowledge & symbols
What was discovered
To reconcile meridian-arc measurements, he proposed choosing the line that minimises the sum of absolute errors subject to the errors summing to zero. It is the first robust data-fitting method and precedes least squares.
It appears in the measurement of a meridian arc in the Papal States with Christopher Maire (1755) and is refined in 1757; Laplace formulated the modern version (1789). There is no disputed priority with Legendre or Gauss, because least squares is later.
What it led to
- Laplace's median and L1-norm method
- robust statistics and quantile regression