st: DHS ten years of data

[Daniel Wilde <dgw24@bath.ac.uk>]

All,

I am trying to create ten years of data for infant and child mortality data 
from a single Demographic Health Survey. I originally had a SAS do file 
which selected the relevant records and variables, and a STATA file that 
then created the two series of DATA. With some help from numerous people on 
STAT list (thanks) I was a able to combine the two do files in a single 
STATA do file. This do file now runs all the way through, but it does not 
always produce tens years of data for both series. For example, for 
Bangledesh 1996 survey (individual recode) it produces ten years of data 
for one series but only seven years for the other series. For Kenya 1989 
(individual recode) it only produces a few years worth of data. Can anyone 
take a look at the do file and tell me why this is happening? I have pasted 
all three do files below:

Any help would be greatly appreciated.

**** DO FILE 1 COMBINED STATA FILE ********


*Explanation of Stata do file and SAS File

set more off

* lets open a log

log using c:\DHSLOG1, replace

* This do file creates ten years worth of time series data for infant, 
under three and child mortality from a demographic health survey.

* This do file has to be run for each survey. Then the results are pasted 
into the panel data time series Excel sheet, which also contains 
information from WDI, then the results are converted back into STATA

* The file is always loaded from the desktop DHSEX

* 	Step 1 Setting up Stata

* The next line tells stata we are using 8.2; because you are using your 
full laptop version of STATA 8, but WED has version 9

version 8.2

* T	he next command increases the systems memory because we are going to be 
working with a lot of variables

set mem 36m

* The next command increases the number of variables it can handle

set matsize 3500

*  Step 2 loading the data

* the next command loads the data file. The file has to be put in

use C:\DHSEX.DTA, clear

* Step 3 Recoding

*Recode the variable that tells you if women interviewed is from a rural or 
urban area

recode v102 2 =  0

*rescale the sample weight

replace v005 = v005/1000000

* Step 4 Extracting the relevant information for each child

* the first stage is too only keep the relevant variables because the 
reshape command crashes if we have everything. We are only keeping 6 kids 
because this is what Stifel SPS prog does, but we need to check that this 
is right with him.

sort caseid
g wid=_n
keep caseid v001 v002 v003 v005 v006 v007 v011 v008 v101 v102 
bidx_01-bidx_06 bord_01-bord_06 b0_01-b0_06   b3_01-b3_06  b4_01-b4_06 
b7_01-b7_06 b10_01- b10_06

* the second stage is too create a macro with all the information we want 
in it apart from the variable that we are making long

global varx "bidx_ bord_ b0_ b3_ b4_ b7_ b10_ "

* the third stage is too reshape so each observation is its own line

reshape long $varx, i(caseid) j(j 01 02 03 04 05 06)

* Lets creat variables with new names

gen hcluster = v001
gen hhnumber = v002
gen  mother   = v003
gen wgt      = v005
gen monthint = v006
gen yearint  = v007
gen dateint  = v008
gen mdob     = v011
gen region   = v101
gen urban    = v102
gen twin     =  b0_
gen dob_      =  b3_
gen sex_      =  b4_
gen aged_     =  b7_
gen flag_     =  b10_

*now lets drop all variables that are not relevant

keep caseid hcluster hhnumber mother wgt monthint yearint dateint mdob 
region urban bidx_ bord_ twin dob_ sex_ aged_

* now lets drop missing observations

drop if bidx == .

* now lets create a variables showing the childs age

gen magekdob = (dob_ - mdob)/12
gen age = dateint - dob
gen yob = int(dob/12)

* now lets sort the variables

sort hcluster hhnumber mother bidx

* from here on is Stifels do file

#delimit;

*Load dataset and keep only those kids born to mothers whose age was 15-39 
at the time of birth;

scalar year = 96;
gen ageint = dateint - dob;
drop if ageint<12;
keep yob aged magekdob wgt ageint;
drop if magekdob<15 | magekdob>35;
save c:\DHSTEMP1, replace;

*------------------------------------------------*
|                                                |
|   Program to create 10 scalars indicating      |
|   the year in which each cohort of kids        |
|   was born.                                    |
|                                                |
|   For example, if the survey year was 1988,    |
|                                                |
|       surv1  = 78                              |
|       surv2  = 79                              |
|         .                                      |
|         .                                      |
|       surv10 = 87                              |
|                                                |
|  Syntax for the program is simply "survgen"    |
|                                                |
*------------------------------------------------*;


capture program drop survgen;
program define       survgen;
  local i = 1;
  while `i' <= 10 {;
    scalar surv`i' = year - 11 + `i';
    local i = `i' + 1;
                  };
  drop if yob == year | yob < surv1;
end;

*------------------------------------------------*
|                                                |
|   Program to calculate IMR & CMR rates         |
|   for cohorts of kids born in each of the      |
|   10 years prior to the survey                 |
|                                                |
|   Saves the output as two datasets:            |
|     c:\DHSTEMP1 -- IMR                       |
|     c:\temp\DHSTEMP2 -- CMR                       |
|                                                |
|   Syntax for the program is simply "cimrgen"   |
|                                                |
*------------------------------------------------*

capture program drop cimrgen;
program define       cimrgen;
 local i = 1;
 local thsnd = 1000;
 qui gen      imrv=0;
  qui replace imrv = 1 if aged  <= 12;
 qui gen      cmrv = 0;
  qui replace cmrv = 1 if aged  <= 36;

sum imrv cmrv;

 while `i' <= 10 {;
    matrix yr`i' = surv`i';
    qui gen x = imrv;
     qui replace x = . if yob ~= surv`i';
    qui summ x [aw=wgt], meanonly;
     local  imrl = r(mean);
     matrix imr`i' = `imrl' * `thsnd';
     drop x;
    if `i' == 2 {;
       matrix imr = ( yr1 , imr1 \ yr2 , imr2 );
                };
    if `i' >  2 {;
       matrix imrn = yr`i' , imr`i';
       matrix imr  = imr \ imrn;
                };

    if `i' <= 8 {;
      qui gen x = cmrv;
       qui replace x = . if yob ~= surv`i';
      qui summ x [aw=wgt], meanonly;
       local  cmrl = r(mean);
       matrix cmr`i' = `cmrl' * `thsnd';
       drop x;

      if `i' == 2 {;
         matrix cmr = ( yr1 , cmr1 \ yr2 , cmr2 );
                  };
      if `i' >  2 {;
         matrix cmrn = yr`i' , cmr`i';
         matrix cmr  = cmr \ cmrn;
                  };
                };

    local i = `i' + 1;
                 };

 matrix colnames imr = year imr;
 matrix colnames cmr = year cmr;
 drop _all;
 svmat cmr, name(col);
 sort year;
 save c:\DHSTEMP1, replace;
 drop _all;
 svmat imr, name(col);
 sort year;
 merge year using c:\DHSTEMP1
 drop _m;
 save c:\DHSTEMP1, replace;
 list;
 graph imr cmr year;

end;

*********************************************;

survgen;
cimrgen;

save c:\DHSEX2;

log close;
clear;


******** DO FILE 2 ORIGINAL SAS FILE 
************************************************

/* SAS program to create SAS dataset of Mortality     */
/* of children in country (cc) in year (YY) from DHS  */

libname user '/home4/ds52/aadata';
libname work1 '/home4/ds52/temp';

options ls=132 ps=54 nocenter;

data work1.temp0;
 set user.bdir3afl;

* Recode urban so that urban = 1 and rural = 0;
if v102 = 2 then v102 = 0;

* Rescale the weights;
v005 = v005/1000000;

********************************************
Extract Info Relevant to Each Child
********************************************;

array bidx_a{20} bidx_01-bidx_20;
array bord_a{20} bord_01-bord_20;
array   b0_a{20}   b0_01-  b0_20;
array   b3_a{20}   b3_01-  b3_20;
array   b4_a{20}   b4_01-  b4_20;
array   b7_a{20}   b7_01-  b7_20;
array  b10_a{20}  b10_01- b10_20;

do i=1 to 6 ;
if bidx_a{i}>0 then do;
 hcluster = v001;
 hhnumber = v002;
 mother   = v003;
 wgt      = v005;
 monthint = v006;
 yearint  = v007;
 dateint  = v008;
 mdob     = v011;
 region   = v101;
 urban    = v102;

 bidx     = bidx_a{i};
 bord     = bord_a{i};
 twin     =   b0_a{i};
 dob      =   b3_a{i};
 sex      =   b4_a{i};
 aged     =   b7_a{i};
 flag     =  b10_a{i};

 keep caseid hcluster hhnumber mother wgt monthint yearint dateint
      mdob region urban bidx bord twin dob sex aged;
if bidx_a{i}>0 then output;
 end;
end;
run;

data work1.temp0;
set work1.temp0;

magekdob = (dob - mdob)/12;
age = dateint - dob;
yob = int(dob/12);
run;

proc sort; by hcluster hhnumber mother bidx; run;
proc contents;
proc means;
weight wgt;
title "Bangladesh 1996 (DHS) Mortality Data of Kids";
run;

libname trn1 xport '/home4/ds52/aadata/bd96mort.v5x';  **;
proc copy in=work1 out=trn1;
select temp0;
run;

endsas;


**** DO FILE THREE --- ORIGINAL STATA DO FILE 
*********************************************

version 6.0
clear
#delimit ;
set matsize 350;
set more off;

log using c:\dstifel\amort\rates\bd96imr.log, replace;

*************************************************************

 Load dataset and keep only those kids born to mothers
 whose age was 15-39 at the time of birth

*************************************************************;

scalar year = 96;

use c:\dstifel\amort\data\bd96mort;
gen ageint = dateint - dob;
drop if ageint<12;
keep yob aged magekdob wgt ageint;
drop if magekdob<15 | magekdob>35;
save c:\temp\temp1, replace;


*------------------------------------------------*
|                                                |
|   Program to create 10 scalars indicating      |
|   the year in which each cohort of kids        |
|   was born.                                    |
|                                                |
|   For example, if the survey year was 1988,    |
|                                                |
|       surv1  = 78                              |
|       surv2  = 79                              |
|         .                                      |
|         .                                      |
|       surv10 = 87                              |
|                                                |
|  Syntax for the program is simply "survgen"    |
|                                                |
*------------------------------------------------*;

capture program drop survgen;
program define       survgen;
  local i = 1;
  while `i' <= 10 {;
    scalar surv`i' = year - 11 + `i';
    local i = `i' + 1;
                  };
  drop if yob == year | yob < surv1;
end;

*------------------------------------------------*
|                                                |
|   Program to calculate IMR & CMR rates         |
|   for cohorts of kids born in each of the      |
|   10 years prior to the survey                 |
|                                                |
|   Saves the output as two datasets:            |
|     c:\temp\temp1 -- IMR                       |
|     c:\temp\temp2 -- CMR                       |
|                                                |
|   Syntax for the program is simply "cimrgen"   |
|                                                |
*------------------------------------------------*

capture program drop cimrgen;
program define       cimrgen;
 local i = 1;
 local thsnd = 1000;
 qui gen      imrv=0;
  qui replace imrv = 1 if aged  <= 12;
 qui gen      cmrv = 0;
  qui replace cmrv = 1 if aged  <= 36;

sum imrv cmrv;

 while `i' <= 10 {;
    matrix yr`i' = surv`i';
    qui gen x = imrv;
     qui replace x = . if yob ~= surv`i';
    qui summ x [aw=wgt], meanonly;
     local  imrl = r(mean);
     matrix imr`i' = `imrl' * `thsnd';
     drop x;
    if `i' == 2 {;
       matrix imr = ( yr1 , imr1 \ yr2 , imr2 );
                };
    if `i' >  2 {;
       matrix imrn = yr`i' , imr`i';
       matrix imr  = imr \ imrn;
                };

    if `i' <= 8 {;
      qui gen x = cmrv;
       qui replace x = . if yob ~= surv`i';
      qui summ x [aw=wgt], meanonly;
       local  cmrl = r(mean);
       matrix cmr`i' = `cmrl' * `thsnd';
       drop x;

      if `i' == 2 {;
         matrix cmr = ( yr1 , cmr1 \ yr2 , cmr2 );
                  };
      if `i' >  2 {;
         matrix cmrn = yr`i' , cmr`i';
         matrix cmr  = cmr \ cmrn;
                  };
                };

    local i = `i' + 1;
                 };

 matrix colnames imr = year imr;
 matrix colnames cmr = year cmr;
 drop _all;
 svmat cmr, name(col);
 sort year;
 save c:\temp\temp1, replace;
 drop _all;
 svmat imr, name(col);
 sort year;
 merge year using c:\temp\temp1;
 drop _m;
 save c:\temp\temp1, replace;
 list;
 graph imr cmr year;

end;

*********************************************;

survgen;
cimrgen;

save c:\dstifel\amort\rates\bd96imr.dta, replace;

log close;
clear;

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/