Python rank: give negative rank to negative numbers-CodePudding

I have a basic set of data like:

ID Value
A  0.1
B  0.2
C  -0.1
D  -0.01
E  0.15

If we use data.rank() we get the result:

ID Value
A  3
B  5
C  1
D  2
E  4

But i want to have so that the negative values result in an negative rank instead such as:

ID Value
A  2
B  3
C  -2
D  -1
E  1

Basically rank give the negative values an negative rank while the postive values get a positive rank but instead of 1 to 5 we get 1 to 3 and -1 to -2. Any help is greatly apreciated.

CodePudding user response：

Rank your positive values and negative values separately and then concat them:

>>> pd.concat([df[df["Value"].gt(0)].rank(),df[df["Value"].lt(0)].mul(-1).rank().mul(-1)]).sort_index()

    ID  Value
0  1.0    1.0
1  2.0    3.0
2 -1.5   -2.0
3 -1.5   -1.0
4  3.0    2.0

CodePudding user response：

Another method similar to the concat answer, but not as compact:

import pandas as pd

A = ['A', 'B', 'C', 'D']
B = [-1, 1, 3, -2]

df = pd.DataFrame({'ID': A, 'value': B})

pos = df.where(df['value'] >= 0)['value'].rank()
neg = df.where(df['value'] < 0)['value'].rank()
pos.update(-neg)

df['rank'] = pos
print(df)

CodePudding user response：

Think outside the box & sort your values.

# create your dummy data
data = pd.DataFrame({'ID':list('ABCDE'), 'Value':[0.1,0.2,-0.1,-0.01,0.15]})

# sort the data so that we can use of the cumsum (and vectorize the operation)
data = data.sort_values('Value')
data['RANK'] = (data['Value']>=0).cumsum()
data['RANK'] -= (data['Value']<0)[::-1].cumsum()

# Sort the values back to the original state ...    
data.sort_values('ID')


     ID     Value   RANK
0     A      0.10    1
1     B      0.20    3
2     C     -0.10   -2
3     D     -0.01   -1
4     E      0.15    2