Converting quarterly to annual data using the first quarter

[Carl Kier]

I know how to convert quarterly to annual data using the collapse command. However, I can't seem to find any examples of converting quarterly data to annual, using the first quarter's data only, which is what I want. Does anyone know how to go about doing that?
Data example (of a limited portion of the data), with the yearly ids generated by me:

* Example generated by -dataex-. For more info, type help dataex
clear
input int omr float(year t) long hus
101 2008 192 26852
101 2008 193 26058
101 2008 194 24502
101 2008 195 22342
101 2009 196 20825
101 2009 197 20847
101 2009 198 21954
101 2009 199 22090
101 2010 200 22775
101 2010 201 24160
101 2010 202 24260
101 2010 203 24357
101 2011 204 23203
101 2011 205 25144
101 2011 206 24307
101 2011 207 22525
101 2012 208 22289
101 2012 209 23782
101 2012 210 23343
101 2012 211 22932
101 2013 212 24217
101 2013 213 24611
101 2013 214 24334
101 2013 215 25883
101 2014 216 25429
101 2014 217 26568
101 2014 218 27162
101 2014 219 27951
101 2015 220 28418
101 2015 221 29930
101 2015 222 30276
101 2015 223 29875
101 2016 224 31561
101 2016 225 31159
101 2016 226 32114
101 2016 227 30752
101 2017 228 32044
101 2017 229 33937
101 2017 230 34306
101 2017 231 34013
101 2018 232 36761
101 2018 233 35334
101 2018 234 35701
101 2018 235 37494
101 2019 236 36347
101 2019 237 36170
101 2019 238 37003
101 2019 239 38894
101 2020 240 37680
101 2020 241 38803
101 2020 242 40458
101 2020 243 42173
101 2021 244 45636
101 2021 245 46966
101 2021 246 49455
101 2021 247 47726
101 2022 248 48374
101 2022 249 50514
101 2022 250 49173
101 2022 251 46075
101 2023 252 44157
101 2023 253 46272
147 2008 192 39382
147 2008 193 36224
147 2008 194 32944
147 2008 195 31799
147 2009 196 24743
147 2009 197 30706
147 2009 198 27445
147 2009 199 26027
147 2010 200 32509
147 2010 201 34462
147 2010 202 32431
147 2010 203 35141
147 2011 204 37246
147 2011 205 37048
147 2011 206 25077
147 2011 207 28405
147 2012 208 27052
147 2012 209 27321
147 2012 210 33137
147 2012 211 34696
147 2013 212 35653
147 2013 213 34015
147 2013 214 30697
147 2013 215 33955
147 2014 216 38057
147 2014 217 35275
147 2014 218 36666
147 2014 219 38240
147 2015 220 39828
147 2015 221 43955
147 2015 222 40725
147 2015 223 43772
147 2016 224 35664
147 2016 225 44624
147 2016 226 42008
147 2016 227 42545
147 2017 228 46969
147 2017 229 43342
end

[Clyde Schechter]

Well, if your data is never missing a first quarter for any omr in any year, you can do this:

sort omr t
collapse (first) hus, by(omr year)

But you should probably verify that assumption before proceeding or you will end up with unwanted second or later quarter results being used instead. So, I would precede the above code with -by omr (t), sort: assert quarter(dofq(t[1])) == 1-.

[Nick Cox]

A little fooling around establishes that a necessary and sufficient test for qdate indicating the first quarter is that mod(qdate, 4) is 0. This also follows directly from the definition that the first quarter of 1960 was 0, so all later and earlier first quarters differ by a multiple of 4.

gen wanted = cond(mod(t, 4) == 0, hus, .) 
collapse wanted, by(omr year)