aggregation in pandas groupby

aggregation in pandas groupby

#create fake data example taken from stackoverflow
df_example = pd.DataFrame({'CG':np.random.randint(0, 5, 100), 'Morph':np.random.choice(['S', 'E'], 100), 'R':np.random.rand(100) * -100})

def my_agg(x):
    x = x.sort_values('R')
    morph = x.head(1)['Morph'].values[0]
    diff = x.iloc[0]['R'] - x.iloc[1]['R']
    diff2 = -2.5*np.log10(sum(10**(-0.4*x['R'])))
    prop = (x['Morph'].iloc[1:] == 'S').mean()
    return pd.Series([morph, diff, diff2, prop], index=['morph', 'diff', 'diff2', 'prop'])

df_example.groupby('CG').apply(my_agg)