Error in Model development

[son1403]

I am trying to calculate ldf & cdf to my development. My triangle show as below:

dev = cl.Development(average = "simple").fit(los_2021) is okay but when i run dev.ldf_, i get the error ValueError: Shape of passed values is (1, 45), indices imply (1, 16)

Could anyone help me with this? Thank you

[kennethshsu]

Hi @son1403,

Can you post the data set here please? Or as an alternative, would you be able to replicate the error using one of the sample datatset included in the package?

[kennethshsu]

I don't think you need to filter your development out of the triangle, that is not the problem. What @jbogaardt was saying is that your triangle is not complete, you might be missing some valuation periods.

Can you try to convert your triangle to OQDQ or OYDY by doing los_2021.gain("OQDQ") and see if that helps?

Also, can you try to call los_2021.origin or los_2021.development to see if you are missing anything?

[son1403]

I tried to convert my Triangle to OQDQ but it doesnt work.

[kennethshsu]

Sorry, it should be los_2021.grain("OQDQ")!

[son1403]

Sorry, it should be los_2021.grain("OQDQ")!

Thank you all @kennethshsu @jbogaardt.
Finally it works.
Appreciate that.

[kennethshsu]

So the problem I think is that like @jbogaardt said, you must be missing some valuation periods.

You can try running los_2021.valuation.unique().sort_values(), it might help you figure out what you are missing:

[Obalus2022]

Hey I am also haveing similar challenges:

I did check valuations, and looks ok,although I have future valuations as well. dont know if that is the problem

[jbogaardt]

@Obalus2022 , this looks odd, but I cannot reproduce. Using the data visible in your screenshot, I get ldfs consistent with the link ratio triangle:

Are you on an older version of chainladder (cl.__version__)?

On the valuation vector, it does include dates for the unfilled lower half of your triangle, so the future dates you are seeing is normal.

reprex code:

import chainladder as cl
import pandas as pd
from io import StringIO
print('chainladder version: ', cl.__version__)

data = """
Idx,Acc,Dev,PDLALAE
PMP,2017,2017,449684
PMP,2018,2018,764067
PMP,2019,2019,428453
PMP,2020,2020,1334759
PMP,2021,2021,1316392
PMP,2017,2018,1485157
PMP,2018,2019,1918224
PMP,2019,2020,2357146
PMP,2020,2021,5503980
PMP,2017,2019,2086793
PMP,2018,2020,2042970
PMP,2019,2021,2871223
PMP,2017,2020,2267049
PMP,2018,2021,2081615
PMP,2017,2021,2444177
"""

Triangles = cl.Triangle(
    pd.read_csv(StringIO(data)),
    origin='Acc', development='Dev', 
    index='Idx', columns='PDLALAE',
    cumulative=True)

dev = cl.Development(average='volume').fit(
    Triangles.loc['PMP', 'PDLALAE', :, :]
)
dev.ldf_