dataframe boolean selection along columns instead of row

Question

Suppose I have the following dataframe:

          a         b         c         d 
0  0.049531  0.408824  0.975756  0.658347
1  0.981644  0.520834  0.258911  0.639664
2  0.641042  0.534873  0.806442  0.066625
3  0.764057  0.063252  0.256748  0.045850

and I want only the subset of columns whose value in row 0 is creater than 0.5. I can do this:

df2 = df.T
myResult = df2[df2.iloc[:, 0] > 0.5].T

But this feels like a horrible hack. Is there a nicer way to do boolean indexing along columns? Somewhere I can specify an axis argument?

I believe you've got the most elegant way out there. – ericmjl Aug 12 '14 at 20:11 — ericmjl, Aug 12 '14 at 20:11

score 7 · Accepted Answer · edited Aug 18 '17 at 18:37

7

How about this?

df.loc[:, df.iloc[0, :] > 0.5]

edited Aug 18 '17 at 18:37

Artur Nowak

5,254
3
22
32

answered Aug 12 '14 at 20:32

chrisb

49,833
8
70
70

Yes, that is precisely what I was looking for. – MorganM Aug 12 '14 at 21:12

score 1 · Answer 2 · answered Aug 12 '14 at 20:32

Another method without using transpose is to create a boolean mask on whether the first row has values larger than 0.5 and then drop the NaN's with a threshold and then finally make a list of the df columns to filter the original df. This is pretty obfuscated though ;)

In [76]:

df[list(df[df.head(1)> 0.5].dropna(thresh=1, axis=1))]
Out[76]:
              c         d
index                    
0      0.975756  0.658347
1      0.258911  0.639664
2      0.806442  0.066625
3      0.256748  0.045850

score 0 · Answer 3 · answered Aug 12 '14 at 20:12

0

Another way of looking at your answer:

In [14]: df.T[df.T[0] > 0.5].T
Out[14]: 
          c        d 
0  0.975756  0.658347
1  0.258911  0.639664
2  0.806442  0.066625
3  0.256748  0.045850

answered Aug 12 '14 at 20:12

ericmjl

13,541
12
51
80

1

Triple transpose might not be as elegant as your answer. – ericmjl Aug 12 '14 at 20:13

dataframe boolean selection along columns instead of row

3 Answers3

Linked