Data handling strategies

• These operations are relatively straightforward in SPSS
• First: Summarising over lower-level units nested within higher
  • individuals within households
  • job or marriage spells within individuals

aggregate out = summary.sav /break = ghid
  /totinc = sum(indinc)
  /number = n.
 . . .
get file summary.sav.
compute aveinc = totinc/number.

AGGREGATE OUTFILE={file} [/MISSING=COLUMNWISE] [/DOCUMENT]
                  {*   }
 [/PRESORTED] /BREAK=varlist[({A})][varlist...]
                              {D}
  /aggvar['label']aggvar['label']...=function(arguments)[/aggvar ...]

SUM     Sum                      MEAN   Mean
SD      Standard deviation       MAX    Maximum
MIN     Minimum                  PGT    % of cases gt value
PLT     % of cases lt value      PIN    % of cases between values
POUT    % of cases not in range  FGT    Fraction gt value
FLT     Fraction lt value        FIN    Fraction between values
FOUT    Fraction not in range    N      Weighted n
NU      Unweighted n             NMISS  Weighted n of missing
NUMISS  Unweighted n of missing  FIRST  First nonmissing
LAST    Last nonmissing

• Stata's equivalent command is collapse:

  . collapse (sum) totinc=ind (count) n=apno, by(ahid)
  . gen aveinc = totinc/n
  

• Second: Mapping higher-level data back to lower levels
• For instance, mapping household level to individual level, either aggregated data like this or data recorded at household level
• To do this we use a many-to-one match, and add data from the household record to every record of an individual in that household

  match files file = individual.sav
             /table = household.sav     /by = hhid.
  

  Both files must be sorted according to hhid.
• Stata syntax:

  . use individual.dta
  . merge hhid using household.dta
  

• Or from scratch, with great efficiency

  . use individual.dta
  . sort ahid apno
  . egen totinc=sum(indinc), by(ahid)
  

• Third: matching individuals across waves - a key panel operation!
• This is a one-to-one match, matching an individual in one file to his/her record in another file.
• SPSS syntax

  match files file = wave1.sav
             /file = wave2.sav   /* "/file" not "/table" */
       /by = pid.
  

  Both files must be sorted according to pid.
• Stata syntax:

  . use wave1.dta
  . merge pid using wave2.dta
  

• Three types of record may result:
  1. with information from wave 1 only
  2. with information from wave 2 only
  3. with information from wave 1 and wave 2
• Stata creates a variable _merge which distinguishes the three cases
• In SPSS one can use the /in= command to create marker variables
• Fourth: matching individuals in households
  • identify individuals with partner in household
    • see wHGSPN: spouse's wPNO, $\leq$ 0 if not present (from the household grid)
  • we thus identify partner's wPNO
  • save WHID, partner's wPNO and substantive variables to a temporary file (renaming variables)
  • merge temporary file back in, linking WHID and wPNO in main file to partner's WHID and wPNO in temporary file
  

  get file = "s:\bhps\spss\aindresp.sav"
        /keep = ahid ahgspn ajbsemp ajbft asex.
  
  select if (ahgspn gt 0).
  sort cases by ahid ahgspn.
  
  save out="m:\spsinfo.sav"
            /rename=(ahgspn ajbsemp ajbft asex
                      = apno aspjbsem aspjbft aspsex).
  

• This is now a file of PIDs and spouses' characteristics: match it back.

  get file = 's:\bhps\spss\aindresp.sav'
    /keep = ahid apno ajbsemp ajbft asex
  match files file = *
             /file = "m:\spsinfo.sav" /by=ahid,apno.
  

• Stata code:

  use ahid ahgspn ajbsemp ajbft asex using s:\bhps\stata\aindresp
  keep  if ahgspn > 0
  rename ahgspn  apno
  rename ajbsemp aspjbsem
  rename ajbft   aspjbft 
  rename asex    aspsex  
  
  sort ahid apno
  
  save m:\spinfo, replace
  
  use ahid apno ahgspn ajbsemp ajbft asex using s:\bhps\stata\aindresp
  sort ahid apno
  merge ahid apno using m:\spinfo
  

• Couples are straightforward because it is a symmetric relationship with only two people; partner's PID is cheap to store
• For more general relationships wEGOALT can be used:
  • A matrix identifying the relationship of every person in the household