I am learning the Concept of Extreme Value Analysis from this guide: RPubs - Extreme Value Analysis
Extreme Value Analysis can be done with following approach which is: Block Maxima, and Peaks Over Threshold, and the Peaks Over Threshold may be done with fitting data to certain thresholds with GPD Distibutions (Generalized Pareto Distribution)
But I cant understand and interpret the Return Level plot means, which is ilustrated like this:
I found some conclusion related to it: "The selection of threshold is not straightforward: threshold too low – bias because of the model asymptotics being invalid; threshold too high – variance is large due to few data points."
is this conclusion based by the dots on return level plot that looks out from the Confidence Interval lines over the years? then which is right in this case? Threshold too low or Threshold too high?
Any Enlightenment with this is really appreciated! Thank You in Advance!