Keeping observations by year

[Budu Gulo]

I am working with three variables company, income, and year (2007 and 2017). How can I keep first 20 observations, based on income (descending order), by year ?

* Example generated by -dataex-. To install: ssc install dataex
clear
input double year str40 company float income
2017 "VN000000TCB8"   6445595
2017 "VN000000VPB6"   6294328
2017 "VN000000BID9"   5122230
2017 "VN000000MBB5"   3519627
2017 "ID1000113707"   3027466
2017 "ID1000092703"   2412458
2017 "ID1000094402"   2175824
2017 "COT29PA00025"   2169839
2017 "COB01PA00030"   2064130
2017 "KR7086790003"   2017741
2017 "ID1000099302"   1804031
2017 "COB51PA00076"   1275266
2017 "ID1000118508"   1220886
2017 "KR7024110009"   1219904
2017 "ID1000123904"   1159370
2017 "VN000000VIB1"   1124279
2017 "COT23PA00010"   1059992
2017 "VN000000TPB0"    938780
2017 "ID1000098205"    748433
2017 "LB0000010415"    747337
2017 "LB0000033441"    726701
2017 "JP3885780001"    603544
2017 "CLP1506A1070"    564815
2017 "JP3699200006"  506691.2
2017 "KR7138930003"  391170.2
2017 "COB23PA00067"    360712
2017 "JP3946750001"    312264
2017 "KR7139130009"    302208
2017 "ID1000128507" 263753.38
2017 "LB0000010530" 235525.05
2017 "COB02PA00012"    219998
2017 "LB0000010613"    200059
2017 "VN000000KLB8"    187711
2017 "KR7175330000" 185062.63
2017 "NGZENITHBNK9"    177614
2017 "JP3117700009"    158455
2017 "ID1000095508" 140495.53
2017 "ID1000103609"    135279
2017 "KR7192530004" 129014.34
2017 "ID1000100407"    121534
2017 "RU000A0JP5V6"    110400
2017 "UG0000000147"    106892
2017 "CLP102411004"    106006
2017 "CL0001692673"    103299
2017 "ID1000107402"     86140
2017 "GB0005405286"  85517.19
2017 "CLP8716M1101"  83134.05
2017 "NGUBA0000001"     76046
2017 "LB0000010332"  70045.25
2017 "ID1000128200" 69497.195
2007 "ZW0009011249"  53725208
2007 "ZW0009011967"  36635152
2007 "ID1000118201"   4838001
2007 "ID1000109507"   4489252
2007 "ID1000095003"   4346224
2007 "KR7105560007"   2757316
2007 "KR7055550008"   2475513
2007 "VN000000VCB4"   2397667
2007 "KR7053000006"   2201994
2007 "ID1000094204"   2116915
2007 "VN000000ACB8"   1732396
2007 "VN000000CTG7"   1149442
2007 "COB07PA00078"   1086923
2007 "KR7004940003"    960945
2007 "ID1000096605"    897928
2007 "ID1000098007"    770481
2007 "ID1000093701"    737905
2007 "JP3500610005"    645233
2007 "ID1000052400"    520719
2007 "VN000000EIB7"    463417
2007 "JP3890350006"    428400
2007 "ID1000093800"    420302
2007 "ID1000115702"    370667
2007 "VN000000HBB5"    365632
2007 "CLP0939W1081"  242287.7
2007 "HU0000061726"    208208
2007 "ID1000109101"  192751.5
2007 "KR7192520005"    160974
2007 "CLP321331116"  135375.8
2007 "VN000000SHB9" 126889.05
2007 "COB52PA00017"    121152
2007 "COB14PA00025"  116555.2
2007 "JP3892100003"    108315
2007 "RU0009029540"    106024
2007 "JP3405000005"    103820
2007 "JP3711200000"     79345
2007 "IS0000001469"     70020
2007 "CNE100000742"     69053
2007 "JP3305990008"     66289
2007 "JP3932800000"     64595
2007 "INE062A01020"   63643.8
2007 "KZ000A0KFFC1"     61354
2007 "CNE000001N05"     56248
2007 "KR7025610007"     54691
2007 "UG0000000386"  53017.36
2007 "KR7007800006"     52797
2007 "ID1000056302"  49554.09
2007 "KR7007200009"     45104
2007 "CNE100000RJ0"     43787
2007 "ID1000055205"  40744.45
end

[Rich Goldstein]

not completely sure I understand, but try this:

gsort year -income
by year: keep if _n<=20

[Budu Gulo]

Thanks Rich Goldstein . The code does the job!