# Selecting threshold for generalized Pareto distribution in R

I'm using the POT Package in R for fitting a Generalized Pareto distribution to my data. For choosing an approximate threshold I'm using the tcplot() and mrlplot() the following way(x is the data containing vector)

tcplot(x,u.range=c(0,quantile(x,probs=0.995)))
mrlplot(x,u.range=c(0,quantile(x,probs=0.995)),col=c("green","black","green"),nt=200 )


But I get a bit confused sometimes as what value I'll choose as threshold. Particularly as mentioned in this site that

interpretation of these plots often requires a good deal of subjective judgement

For example what would be the threshold given the following plot?

-

## 1 Answer

I believe the threshold is selected as the point at which the MRL plot becomes kind of linear, which to my eye is 2.75 on the MRL plot.

-