Largest five numbers when there are three columns

March 1, 2022

a) How to find the largest five SNRs when the ID is repeated? And also I want all these three columns as the output.
b) I also want the eliminated lines as the output.

```
      FITS             ID              SNR
  1004234.fits   J16355032-2814188     714
  1004235.fits   J16355032-2814188     444
  1004236.fits   J16355032-2814188     331
  1004237.fits   J16355032-2814188     492
  1004238.fits   J16355032-2814188     690
  1004239.fits   J16355032-2814188     491
  1004240.fits   J16355032-2814188     489
  1004241.fits   J16355032-2814188     382
  1004242.fits   J16355032-2814188     635
  1004243.fits   J16355032-2814188     522
  1004244.fits   J16355032-2814188     385
  1004245.fits   J16355032-2814188     645
  1004475.fits   J22152631+4958343      62
  1004476.fits   J22152631+4958343     162
  1004477.fits   J22152631+4958343      76
  1004478.fits   J22152631+4958343     103
  1005113.fits   J22154748+4954052     212
  1005114.fits   J22154748+4954052     227
  1005115.fits   J22154748+4954052     148
  1005116.fits   J22154748+4954052     160´´´

This is the expected output for a.

```      FITS             ID              SNR
   1004234.fits   J16355032-2814188     714
   1004235.fits   J16355032-2814188     690
   1004236.fits   J16355032-2814188     645
   1004237.fits   J16355032-2814188     635
   1004238.fits   J16355032-2814188     522
   1004475.fits   J22152631+4958343      62
   1004476.fits   J22152631+4958343     162
   1004477.fits   J22152631+4958343      76
   1004478.fits   J22152631+4958343     103
   1005113.fits   J22154748+4954052     212
   1005114.fits   J22154748+4954052     227
   1005115.fits   J22154748+4954052     148
   1005116.fits   J22154748+4954052     160  ´´´

Expected output for b.

      ``` FITS             ID              SNR

    1004235.fits   J16355032-2814188     444
    1004236.fits   J16355032-2814188     331
    1004237.fits   J16355032-2814188     492
    1004239.fits   J16355032-2814188     491
    1004240.fits   J16355032-2814188     489
    1004241.fits   J16355032-2814188     382
    1004243.fits   J16355032-2814188     522
    1004244.fits   J16355032-2814188     385  ´´´

>Solution :

Use groupby+head on the sorted dataframe to get the indices, then slice:

idx = df.sort_values(by='SNR', ascending=False).groupby('ID').head(5).index

df2 = df.loc[idx]

output:

            FITS                 ID  SNR
0   1004234.fits  J16355032-2814188  714
4   1004238.fits  J16355032-2814188  690
11  1004245.fits  J16355032-2814188  645
8   1004242.fits  J16355032-2814188  635
9   1004243.fits  J16355032-2814188  522
17  1005114.fits  J22154748+4954052  227
16  1005113.fits  J22154748+4954052  212
13  1004476.fits  J22152631+4958343  162
19  1005116.fits  J22154748+4954052  160
18  1005115.fits  J22154748+4954052  148
15  1004478.fits  J22152631+4958343  103
14  1004477.fits  J22152631+4958343   76
12  1004475.fits  J22152631+4958343   62

Other rows:

df3 = df.loc[df.index.difference(idx)]

output:

            FITS                 ID  SNR
1   1004235.fits  J16355032-2814188  444
2   1004236.fits  J16355032-2814188  331
3   1004237.fits  J16355032-2814188  492
5   1004239.fits  J16355032-2814188  491
6   1004240.fits  J16355032-2814188  489
7   1004241.fits  J16355032-2814188  382
10  1004244.fits  J16355032-2814188  385