I have a problem with selecting a threshold for a dataset I cannot show here. In a MRL-plot, a suitable threshold should be linear with the gradient ξ/(1−ξ) and intercept σ_u/(1−ξ) (as of my understanding).
And a parameter stability plot
Based on the parameter stability-plot a stable threshold would be around 0.09-0.10. And if I use the coefficiants of the fitted GDP based on a threshold of 0.09, I get the following "line" in the MRL-plot:
Is this a suitable threshold?