Pandas: aggregating multiple columns with multiple functions

Question

Pandas in Python and Dplyr in R are both flexible data wrangling tools. For example, in R, with dplyr one can do the following;

custom_func <- function(col1, col2) length(col1) + length(col2)

ChickWeight %>% 
  group_by(Diet) %>% 
  summarise(m_weight = mean(weight), 
            var_time = var(Time), 
            covar = cov(weight, Time),
            odd_stat = custom_func(weight, Time))

Notice how in one statement;

• I can aggregate over multiple columns in one line.
• I can apply different functions over these multiple columns in one line.
• I can use functions that take into account two columns.
• I can throw in custom functions for any of these.
• I can declare new column names for these aggregations.

Is such a pattern also possible in pandas? Note that I am interested in doing this in a short statement (so not creating three different dataframes and then joining them).

Answer 1

With pandas groupby.apply() you can run multiple functions in a groupby aggregation. Please note for statistical functions you would need scipy installed. For custom functions will need to run an aggregate like sum() for groupwise data:

def customfct(x,y):
    data = x / y
    return data.mean()

def f(row):  
    row['m_weight'] = row['weight'].mean()
    row['var_time'] = row['Time'].var()
    row['cov'] = row['weight'].cov(row['Time'])
    row['odd_stat'] = customfct(row['weight'], row['Time'])
    return row

aggdf = df.groupby('Diet').apply(f)