ImplicitDifferentiables.jl

When estimating models economics, conceptually they can be can be thought of as a mapping f from a set of parameters p to some outcome that is a real number, eg a log likelihood or a moment distance.

However, technically f may involve

1. A mapping from a vector of real numbers c that is mapped to p, so that one can use an unconstrained or constrained generic optimizer or MCMC algorithm.

2. Solving for some implicit quantity y, eg equlibrium prices, that satisfies an equation g(y, p). This is usually costly and involves a numerical rootfinder.

This is a lightweight interface package that helps dealing with these situations. Please read the manual for details.

The package is still experimental. Interface changes will respect SemVer, but that's all.