Change the grain of my data

[Coomaren]

This is the first 15 rows of my data

Origin | Month | Value | OriginDate | Develop
2012 | 3 | NaN | 01/01/2012 | 01/04/2012
2012 | 6 | NaN | 01/01/2012 | 01/07/2012
2012 | 9 | 0.000 | 01/01/2012 | 01/10/2012
2012 | 12 | 0.000 | 01/01/2012 | 01/01/2013
2012 | 15 | 214.685 | 01/01/2012 | 01/04/2013
2012 | 18 | 1075.115 | 01/01/2012 | 01/07/2013
2012 | 21 | 1199.062 | 01/01/2012 | 01/10/2013
2012 | 24 | 1735.588 | 01/01/2012 | 01/01/2014
2012 | 27 | 2374.559 | 01/01/2012 | 01/04/2014
2012 | 30 | 2523.518 | 01/01/2012 | 01/07/2014
2012 | 33 | 2734.180 | 01/01/2012 | 01/10/2014
2012 | 36 | 2746.981 | 01/01/2012 | 01/01/2015
2012 | 39 | 3509.450 | 01/01/2012 | 01/04/2015
2012 | 42 | 3873.238 | 01/01/2012 | 01/07/2015
2012 | 45 | 5045.910 | 01/01/2012 | 01/10/2015

I create the triangle and the grain is OYDY

ConstTrian = cl.Triangle(
Construction,
origin="OriginDate",
development="Develop",
columns="Value",
cumulative=True,
).dropna()

I tried to change the grain back to OYDQ through

ConstTrian.grain('OYDQ')

The error is "ValueError: New grain not compatible with existing grain", which is strange since my data development was originally in quarter

Any help on either loading the data or changing the grain is most welcome.

[kennethshsu]

Your dates format are a bit unusual so I wonder if python is confused. I would recommend explicitly specifying the date format first.

df = pd.read_csv("/Users/kennethhsu/Desktop/test.csv")
df["OriginDate"] = pd.to_datetime(df["OriginDate"], format="%d/%m/%y")
df["Develop"] = pd.to_datetime(df["Develop"], format="%d/%m/%y")

Next, the valuation date should be the end of the month. So do this:

df["Valuation"] = df["Develop"] - pd.Timedelta(days=1)
df

Now you are ready to go into chainladder:

triangle = cl.Triangle(
    df,
    origin="OriginDate",
    development="Valuation",
    columns="Value",
    cumulative=True,
    grain="OYDQ"
)
triangle

Can you try the above and see if it works for you?

test.csv

[Coomaren]

I forced the dates to be recognised as dates by loading through Excel. The development was automatically recognised as quarters. Thank you.