Using an array of ELRs for Born-Ferg

[Obalus2022]

Hi again another question. Reviewing how chainladder handles BF. can I use a array of ELRs, or do I have to just use one?

I tried it this way and it didnt work

[kennethshsu]

I think you would have to modify your sample_weight directly, as the apriori needs to be one single hyperparameter.

You can do something like this, but it's not elegant:

raa = cl.load_sample("raa")
apriori = raa.latest_diagonal * 0 + 50000
apriori.loc[:, :, "1981",] = apriori.loc[:, :, "1981",] * 0.7
apriori.loc[:, :, "1982",] = apriori.loc[:, :, "1982",] * 0.6
cl.BornhuetterFerguson(apriori=1.0).fit(X=raa, sample_weight=apriori).ultimate_

Maybe @jbogaardt knows a better way