changing sort in value_counts

Question

If I do

mt = mobile.PattLen.value_counts()   # sort True by default 

I get

4    2831 3    2555  5    1561 [...] 

If I do

mt = mobile.PattLen.value_counts(sort=False)  

I get

8    225 9    120 2   1234  [...] 

What I am trying to do is get the output in 2, 3, 4 ascending order (the left numeric column). Can I change value_counts somehow or do I need to use a different function.

Answer 1

I think you need sort_index, because the left column is called index. The full command would be mt = mobile.PattLen.value_counts().sort_index(). For example:

mobile = pd.DataFrame({'PattLen':[1,1,2,6,6,7,7,7,7,8]}) print (mobile)    PattLen 0        1 1        1 2        2 3        6 4        6 5        7 6        7 7        7 8        7 9        8  print (mobile.PattLen.value_counts()) 7    4 6    2 1    2 8    1 2    1 Name: PattLen, dtype: int64   mt = mobile.PattLen.value_counts().sort_index() print (mt) 1    2 2    1 6    2 7    4 8    1 Name: PattLen, dtype: int64 

Answer 2

As hinted by normanius’ comment under jezrael’s answer :

>>> df = pd.DataFrame({"a":[1,1,2,6,6,7,7,7,7,8]}) >>> df.a.value_counts()[df.a.unique()] 1    2 2    1 6    2 7    4 8    1 Name: a, dtype: int64 

one can sort by any order by providing a custom index explicitely :

>>> df.a.value_counts()[[8,7,6,2,1]] 8    1 7    4 6    2 2    1 1    2 Name: a, dtype: int64 >>> df.a.value_counts()[[1,8,6,2,7]] 1    2 8    1 6    2 2    1 7    4 Name: a, dtype: int64 

This is of particular interest for plotting categorical data :

>>> df.a.value_counts()[['hourly','daily','weekly','monthly']].plot(type="bar") 

Anecdotically, it can be used to remove some entries or to make others appear several times :

>>> df.a.value_counts()[[1,1,1,8]] 1    2 1    2 1    2 8    1 Name: a, dtype: int64