fit_model returning NaN

[kharelg1]

I am getting "-inf" for the return value and NaN for the confidence intervals after running the fit_model() - see attached file (issue1).

I used the following script for the attached data:
model = EVA(df)
model.from_extremes(df, method='BM',extremes_type='high', block_size='365.2425D')
model.set_extremes(df)
model.plot_extremes(figsize=(8,5), ax=None, show_clusters=False)
fitting = model.fit_model(model='MLE', distribution='genextreme')
print (fitting)
r_period = [1,25,100,500,1000]
summary = model.get_summary(r_period, return_period_size='365.2425D', alpha=0.95, n_samples=10000)
print (summary)
test_annual_max_series.txt

I would appreciate it if you could help in finding anything wrong with the script and the output.

[georgebv]

This is expected when using block size of 1 year and looking for 1-year return value. Effectively, you are asking a question of "what is the value which has a 100% probability of being exceeded". An answer to that is naturally negative infinity. I recommend reading about GEVD to understand why this is happening.