Finding the first occurrence per index in Multiindex dataframe

Question

I have a pandas dataframe called panel_long looking something like that:

                 x  y  z event Start
company period   
A       1        1  2  3 False 5
B       1        1  2  3 False 7
...
A       27       1  2  3 True  5
B       27       1  2  3 False 7
...
A       31       1  2  3 False 5
B       31       1  2  3 True  7

As a preparation for a survival analysis with livelines I'd like to add a column which states the first Period event equals True per company. For non-multiindex dataframes I'd use idxmax, but I couldn't figure out how to make it work for a multiindex df. Among others, I've tried the following:

def funct(df, column):
    df[column].idxmax()

df = panel_long.groupby('company').apply(funct, ('event'))

But that just returns an empty dataframe. Any ideas? I'm stuck

Answer 1

You should return the output in your self-define function

def funct(df, column):
    return df[column].idxmax()
df.groupby('company').apply(funct, ('event'))
Out[248]: 
company
A    (A, 1)
B    (B, 1)
dtype: object