在 Pandas GroupBy object 中减去两列

Question

I have a dataset with marketing campaigns, where each house receive campaign actions like "flyer", or "call". Each action has it's own creation and end date. Some houses have only 1 action, and some have a couple.

What I want to do is:

I want to calculate the length of the campaign for each house, so the time between the first action (eg flyer) and the last recorded action for each house. If each house had only 1 action, I could easily solve this by subtracting the end date column with the start date column.

Because some houses have multiple actions, I figured I could group all the houses with the Pandas GroupBy function. Does anyone know how to subtract within a groupby object?

Data looks like this:

house1 flyer 01-12-2014 05-12-2014
house1 phonecall 06-12-2014 06-12-2014
house2 flyer 01-12-2014 31-12-2014

my expected output looks like this:

house1 ; 5 days
house2 ; 30 days
house3 ; 12 days
house4 ; 60 days
etc

Answer 1

Simply use the agg function on groups:

t = df.groupby("house").agg({"start": min, "end": max})
t["duration"] = t.end - t.start

The result is:

            start        end duration
house                                
house1 2014-01-12 2014-06-12 151 days
house2 2014-01-12 2014-12-31 353 days

Edit - creating the dataframe

Per a question in one of the comments, here's how I created the dataframe:

data = """house1 flyer 01-12-2014 05-12-2014
house1 phonecall 06-12-2014 06-12-2014
house2 flyer 01-12-2014 31-12-2014"""

df = pd.read_csv(StringIO(data), sep = "\s+", 
                 header = None, 
                 names = ["house", "medium", "start", "end"])

# Make sure 'start' and 'end' are dates. 
df.end = pd.to_datetime(df.end)
df.start = pd.to_datetime(df.start)