MetalogOLSSolver

Solves the metalog least squares problem.

Because a metalog quantile function is linear in its coefficients, fitting one is ordinary least squares and needs no iteration. The solve goes through a QR decomposition rather than the normal equations, which squares the condition number and so loses roughly half the available precision as the number of terms grows.

An optional ridge term guards against the ill-conditioning that appears with many terms, at the cost of shrinking the coefficients slightly toward zero.

Parameters