Closed OliverFHD closed 4 years ago
This problem goes away once phi(tau) is expressed as
phi(tau) = lambda(tau)*dphi/dlambda(tau) - 0.5 tau^2 .
This is because the polynomial coefficients of dphi/dlambda(tau) can be fit much more robustly (than those of phi(tau)).
When fitting the cumulant generating function with polynomials of very high order in tau (order = 19) , the inverse Laplace transform can get stuck in the high density tail.