--------------------------------------------------------------------------------------------------------------------------------------------------------------------
      name:  <unnamed>
       log:  e:\DHS\programs\tests\diffs_between_surveys_log_22July2016.txt
  log type:  text
 opened on:  25 Jul 2016, 08:28:09

. 
. * Tom Pullum, tom.pullum@icfi.com, July 25, 2016
. 
. set more off

. cd e:\DHS\DHS_data\KR_files
e:\DHS\DHS_data\KR_files

. 
. ****************************************************************
. 
. program define setup1
  1. 
. * Construct the indicator, number the surveys, save the needed variables
. 
. scalar ssurvey=ssurvey+1
  2. local lsurvey=ssurvey
  3. gen survey=ssurvey
  4. 
. * CONSTRUCT THE INDICATOR
. 
. * values other than 0 and 1 should be interpreted as .
. replace g100=. if g100>1
  5. replace g102=. if g102>1
  6. 
. gen y = .
  7. replace y=0 if g100<.
  8. replace y=1 if g102==1  
  9. 
. keep v005 v021 v023 y survey
 10. save temp_`lsurvey'.dta, replace
 11. 
. end

. 
. ****************************************************************
. 
. program define setup2
  1. 
. * Combine the surveys into one file
. 
. use temp_1.dta, clear
  2. append using temp_2.dta
  3. 
. egen cluster=group(v021 survey) 
  4. egen stratum=group(v023 survey) 
  5. 
. save temp.dta, replace
  6. 
. end

. 
. ****************************************************************
. 
. program define analyze
  1. 
. * Test whether the "survey" variable statistically significant
. 
. svyset cluster [pweight=v005], strata(stratum) singleunit(scaled)
  2. 
. tab survey y
  3. tab survey y [iweight=v005/1000000], row
  4. 
. * Test for significance of change or difference
. svy: logit y i.survey
  5. scalar p=e(p)
  6. scalar list p
  7. 
. end

. 
. ****************************************************************
. ****************************************************************
. ****************************************************************
. ****************************************************************
. ****************************************************************
. * EXECUTION BEGINS HERE
. 
. * Example:  difference between two surveys in FGM prevalence
. 
. * Kenya 27.1% in 2008-09 vs 21.0% in 2014
. 
. * Uganda 0.6% in 2006 vs. 1.4% in 2011
. 
. scalar ssurvey=0

. 
. use e:\DHS\DHS_data\IR_files\KEIR52FL.dta, clear 

. setup1
(7 real changes made, 7 to missing)
(4 real changes made, 4 to missing)
(8,444 missing values generated)
(8,437 real changes made)
(2,541 real changes made)
file temp_1.dta saved

. 
. use e:\DHS\DHS_data\IR_files\KEIR70FL.dta, clear 

. setup1
(0 real changes made)
(0 real changes made)
(31,079 missing values generated)
(14,739 real changes made)
(4,377 real changes made)
file temp_2.dta saved

. 
. setup2
file temp.dta saved

. analyze

      pweight: v005
          VCE: linearized
  Single unit: scaled
     Strata 1: stratum
         SU 1: cluster
        FPC 1: <zero>

           |           y
    survey |         0          1 |     Total
-----------+----------------------+----------
         1 |     5,896      2,541 |     8,437 
         2 |    10,362      4,377 |    14,739 
-----------+----------------------+----------
     Total |    16,258      6,918 |    23,176 


+----------------+
| Key            |
|----------------|
|   frequency    |
| row percentage |
+----------------+

           |           y
    survey |         0          1 |     Total
-----------+----------------------+----------
         1 | 6,154.278  2,284.252 |  8,438.53 
           |     72.93      27.07 |    100.00 
-----------+----------------------+----------
         2 | 11,557.23  3,065.936 |14,623.162 
           |     79.03      20.97 |    100.00 
-----------+----------------------+----------
     Total |  17,711.5  5,350.188 | 23,061.69 
           |     76.80      23.20 |    100.00 

(running logit on estimation sample)

Survey: Logistic regression

Number of strata   =       100             Number of obs     =          23,176
Number of PSUs     =     1,991             Population size   =  23,061,691,753
                                           Design df         =           1,891
                                           F(   1,   1891)   =           10.60
                                           Prob > F          =          0.0011

------------------------------------------------------------------------------
             |             Linearized
           y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    2.survey |   -.335849    .103141    -3.26   0.001    -.5381312   -.1335669
       _cons |  -.9911088   .0958041   -10.35   0.000    -1.179002    -.803216
------------------------------------------------------------------------------
         p =  .00114906

. 
. 
. clear

. 
. scalar ssurvey=0

. 
. * In the Uganda surveys we must construct the strata and rename the outcome variables
. 
. * Uganda 2006
. use e:\DHS\DHS_data\IR_files\UGIR52FL.dta

. rename s643a g100

. rename s643b g102

. drop v023

. egen v023=group(v024 v025)

. setup1
(1 real change made, 1 to missing)
(13 real changes made, 13 to missing)
(8,531 missing values generated)
(8,530 real changes made)
(61 real changes made)
file temp_1.dta saved

. 
. * Uganda 2011
. use e:\DHS\DHS_data\IR_files\UGIR60FL.dta 

. rename s631d g100

. rename s631f g102

. drop v023

. egen v023=group(v024 v025)

. setup1
(33 real changes made, 33 to missing)
(3 real changes made, 3 to missing)
(8,674 missing values generated)
(8,641 real changes made)
(158 real changes made)
file temp_2.dta saved

. 
. setup2
file temp.dta saved

. analyze

      pweight: v005
          VCE: linearized
  Single unit: scaled
     Strata 1: stratum
         SU 1: cluster
        FPC 1: <zero>

           |           y
    survey |         0          1 |     Total
-----------+----------------------+----------
         1 |     8,469         61 |     8,530 
         2 |     8,483        158 |     8,641 
-----------+----------------------+----------
     Total |    16,952        219 |    17,171 


+----------------+
| Key            |
|----------------|
|   frequency    |
| row percentage |
+----------------+

           |           y
    survey |         0          1 |     Total
-----------+----------------------+----------
         1 | 8,475.667  54.700871 |8,530.3683 
           |     99.36       0.64 |    100.00 
-----------+----------------------+----------
         2 | 8,514.466 123.720288 | 8,638.186 
           |     98.57       1.43 |    100.00 
-----------+----------------------+----------
     Total | 16,990.13 178.421159 | 17,168.55 
           |     98.96       1.04 |    100.00 

(running logit on estimation sample)

Survey: Logistic regression

Number of strata   =        36             Number of obs     =          17,171
Number of PSUs     =       772             Population size   =  17,168,554,314
                                           Design df         =             736
                                           F(   1,    736)   =            4.56
                                           Prob > F          =          0.0331

------------------------------------------------------------------------------
             |             Linearized
           y |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
-------------+----------------------------------------------------------------
    2.survey |   .8115765   .3802565     2.13   0.033     .0650597    1.558093
       _cons |  -5.043075   .3438202   -14.67   0.000     -5.71806    -4.36809
------------------------------------------------------------------------------
         p =  .03314916

. 
. 
. * p is the significance of a test of H0: in the population, there was no difference 
. *  in the prevalence of the outcome across the surveys
. 
. 
end of do-file

. exit, clear