Derivatives

A central method in Fisher Forecast is the numerical derivatives of observables. However, since the Angular Coefficients are evaluated performing numerical integration of solution of the Einstein-Boltzmann equations, finite-differences methods are quite unstable. Up to now, we have included the following methods

CosmoCentral.SteMDerivative — Function

SteMDerivative(x::Vector{Float64}, y::Vector{Float64})

This function evaluates the numerical derivative according to the SteM algorithm, Camera et al. 2017:

a linear regression over $x$ and $y$ is performed
if the points obtained with the fit are close enough (less than 0.01 relative difference) the linear ansatz is satisfied and the slope gives the derivative
if the linear ansatz is not satisfied, the external couple of points and the linear regression is performed again till the linear ansatz is satisfied

source