Sorting Multi level dataframe index level0, based on value of index level1 at specific column

Question

I have this multi level dataframe

import pandas as pd
sites = pd.Index(['CCC', 'RRR','TTT', 'SSS'], name='SITELIST')
vvv = pd.DataFrame({'KK':[1,2,3,4],'GG':[4,3,4,5], 'RR':[1,5,1,8]}, index = sites)
ttt = pd.DataFrame({'KK':[11,22,33,34],'GG':[23,33,34,35], 'RR':[1,2,2,1]}, index = sites)
df = pd.concat([vvv, ttt], axis=0, keys=["VVV", "TTT"], names=['Sensor', 'Site'])
df = df.swaplevel().sort_index(level=0).reindex(['VVV','TTT'],level=1)
print(df)

dataframe :

             KK  GG  RR
Site Sensor            
CCC  VVV      1   4   1
     TTT     11  23   1
RRR  VVV      2   3   5
     TTT     22  33   2
SSS  VVV      4   5   8
     TTT     34  35   1
TTT  VVV      3   4   1
     TTT     33  34   2

How do i sort the dataframe index level0 'Site' using each value on index level1 'VVV' at column RR The expected result as such:

             KK  GG  RR
Site Sensor            
CCC  VVV      1   4   1 <-
     TTT     11  23   1
TTT  VVV      3   4   1 <-
     TTT     33  34   2
RRR  VVV      2   3   5 <-
     TTT     22  33   2
SSS  VVV      4   5   8 <-
     TTT     34  35   1

I have tried to sort with

print(df.sort_values('RR'))

But get unexpected

             KK  GG  RR
Site Sensor            
CCC  VVV      1   4   1
     TTT     11  23   1
SSS  TTT     34  35   1
TTT  VVV      3   4   1
RRR  TTT     22  33   2
TTT  TTT     33  34   2
RRR  VVV      2   3   5
SSS  VVV      4   5   8

Thank you.

I need the index level0 which is 'Site' is sorted by the value of col RR but on;y the one in index level1 VVV (or each row VVV). If I just sort_value the col RR; then value on row TTT will also be sorted, thus create something such as: print(df.sort_values('RR')) — Bangkit Cipta Persada
– Bangkit Cipta Persada, Commented Sep 20, 2020 at 18:31

Henry Yik · Accepted Answer · 2020-09-20 18:52:14Z

2

IIUC you can first reset_index to sort and get the correct order of site:

df = (df.reset_index().sort_values(["Sensor", "RR", "Site"], ascending=(False, True, True)))

   Site Sensor  KK  GG  RR
0  CCC    VVV   1   4   1
6  TTT    VVV   3   4   1
2  RRR    VVV   2   3   5
4  SSS    VVV   4   5   8
1  CCC    TTT  11  23   1
5  SSS    TTT  34  35   1
3  RRR    TTT  22  33   2
7  TTT    TTT  33  34   2

From here using the result to construct an ordered pd.Categorical so you can obtain the final result:

df["Site"] = pd.Categorical(df["Site"], categories=df["Site"].unique(), ordered=True)

print (df.sort_values("Site").set_index(["Site","Sensor"]))

             KK  GG  RR
Site Sensor            
CCC  VVV      1   4   1
     TTT     11  23   1
TTT  VVV      3   4   1
     TTT     33  34   2
RRR  VVV      2   3   5
     TTT     22  33   2
SSS  VVV      4   5   8
     TTT     34  35   1

answered Sep 20, 2020 at 18:52

Henry Yik

22.6k5 gold badges21 silver badges44 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Bangkit Cipta Persada Over a year ago

I was thinking it should just be as df.sort_index('Site', by=[index['Sensor'], col['RR']]) should be more intuitive :)

Collectives™ on Stack Overflow

Sorting Multi level dataframe index level0, based on value of index level1 at specific column

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related