why groupby.apply return duplicate level

Question

def afun(group):
  aa=len(group)
  group.sort_values(inplace=True)

  return pandas.DataFrame({'score':numpy.arange(aa),'price':group})



df = pandas.DataFrame({
              'stock':numpy.repeat( ['AAPL','GOOG','YHOO'], 3 ),
              'date':numpy.tile( pandas.date_range('5/5/2015', periods=3, freq='D'), 3 ),
              'price':(numpy.random.randn(9).cumsum() + 10) ,
              'price2':(numpy.random.randn(9).cumsum() + 10)})
df = df.set_index(['stock','date'])
agroupDf=df.groupby(level='date') 



tt=agroupDf['price'].apply(afun)

the value of variable tt is shown in the figure

my question is why tt has two columns 'date',and how to avoid the secend column 'date'?

Answer 1

The first one is the groupby 'date'. The second one is the index 'date'.

changing things around - this time groupby stock:

df       = df.set_index(['date','stock'])
agroupDf = df.groupby(level='stock') 

tt=agroupDf['price'].apply(afun)
tt


                            price  score
stock date       stock                  
AAPL  2015-05-05 AAPL    9.333143      0
      2015-05-06 AAPL    9.680022      1
      2015-05-07 AAPL    9.870889      2
GOOG  2015-05-06 GOOG   10.030032      0
      2015-05-05 GOOG   10.229084      1
      2015-05-07 GOOG   10.571631      2
YHOO  2015-05-07 YHOO    9.996925      0
      2015-05-05 YHOO   10.342180      1
      2015-05-06 YHOO   10.586120      2

I think you want this:

df       = df.set_index('stock')
agroupDf = df.groupby('date') 

tt=agroupDf['price'].apply(afun)
tt

                          price  score
date       stock                  
2015-05-05 AAPL   10.414396      0
           GOOG   12.608225      1
           YHOO   12.830496      2
2015-05-06 AAPL   10.428767      0
           GOOG   11.189663      1
           YHOO   11.988177      2
2015-05-07 YHOO   11.202677      0
           AAPL   11.274440      1
           GOOG   11.780654      2