* Construction of table 4.4 (marriage before age 18) in the Bangladesh 2017 final report

* General strategy for EMW surveys: for each original case, add a second case
*  with residual weight and never-married status

* Specify a workspace
cd e:\DHS\DHS_data\scratch

* Read the IR file
use "...BDIR7RFL.DTA", clear 

* Must match the covariates in the table with the correct version of awfact
* Total:                 awfactt
* Residence:       v025, awfactu
* Division:        v024, awfactr
* Education:       v149, awfacte
* Wealth quintile: v190, awfactw

local letters t u r e w

* restrict to age 20-24
keep if v013==2

keep v001 v002 v003 v005 v024 v025 v149 v190 v511 awfact*
gen EMW=1
gen afm=v511

* Construct weights wt* for the original EMW cases
foreach ll of local letters {
gen wt`ll'=v005
}

save EMW.dta, replace

replace EMW=0
replace afm=99

* Construct corresponding weights wt* for the artificial NMW cases
foreach ll of local letters {
replace wt`ll'=int(((awfact`ll'-100)/100)*v005)
}

* Combine the EMW and NMW cases
quietly append using EMW.dta

* Construct the outcome, married before age 18
gen by18=0
replace by18=100 if afm<18

*save ALL.dta, replace

* Table 4.4. Marriage before age 18
* Note; the %'s and n's are produced separately.
* Ignore (!!) the totals rows for the separate panels.
  
* Total
summarize by18 [iweight=wtt/1000000]

* Residence
tab v025 [fweight=wtu], summarize(by18) means
tab v025 [iweight=wtu/1000000]

* Division
tab v024 [fweight=wtr], summarize(by18) means
tab v024 [iweight=wtr/1000000]

* Education
tab v149 [fweight=wte], summarize(by18) means
tab v149 [iweight=wte/1000000]

* Wealth quintile
tab v190 [fweight=wtw], summarize(by18) means
tab v190 [iweight=wtw/1000000]