Displaying 1 result from an estimated 1 matches for "log_sigma".
2008 Oct 23
1
distribution fitting
...right?
The real problems come out when I try to fit a General Pareto distribution.
Using the same method, I calculated the negative log likelihood function for
the General Pareto distribution:
> # negative log-likelihood function for general pareto distribution
> nll_gpd <- function(xi, log_sigma, mu) { # shape, scale, location
+ n <- length(interarrival_times)
+ x <- interarrival_times
+ cat("xi:",xi,"; log_sigma:",log_sigma,"; mu:",mu,"\n")
+ n*log_sigma + (xi+1)*sum(log(1+xi*(x-mu)/exp(log_sigma)))/xi
+ }
and...