Usage of cl.Development() and transform

[cooneycw]

I am using the following code to create ldf's via the cl.Development() method:

model = cl.Development(average='simple').fit(triangle_df)

My understanding is that i can then use:

ult = model.transform(triangle_df)

to apply the ldf's and extend the triangle to ultimate via the factors. This step executes without any message. Next, when i attempt:

ult.full_triangle_

however, i am experiencing the following error:

AttributeError: 'Triangle' object has no attribute 'full_triangle_'

ult does show a triangle of data with NaN values equivalent to triangle_df. Can someone pls help me understand what i am doing wrong? I have also tried using:

cl.Development(average = 'simple').fit_transform(triangle_df)

With the same outcome. Many thanks.

[jbogaardt]

Development only produces development patterns, to get ultimates and a full triangle, you'd need to additionally fit one of the IBNR models (Chainladder, BornhuetterFerguson, etc).

dev = cl.Development(average='simple').fit_transform(triangle_df)
model = cl.Chainladder().fit(dev)

or more compactly:

cl.Pipeline(
    [('dev', cl.Development(average='simple')),
     ('model', cl.Chainladder())]
).fit(triangle_df)

[cooneycw]

Thank you. And were i to make adjustments to my ldf's via drop, etc., these would then flow through into the Chainladder step?

[jbogaardt]

Correct. Feel free to check this by accessing the ldf_ property on the chainladder step to verify.

pipe = cl.Pipeline(
    [('dev', cl.Development(average='simple')),
     ('model', cl.Chainladder())]
).fit(triangle_df)
print(pipe.named_steps.model.ldf_)
print(pipe.named_steps.dev.ldf_)