/*--------------------------------------------------- HOW TO USE THIS FILE This is an HTML version of the Stata do file ex3.do It is intended to show you the code and to allow links not to use interactively use ex3.do in the data editor for this EVERYTHING INSIDE (STAR SLASH) AND (SLASH STAR) IS TAKEN AS A COMMENT These are shown in black here ---------------------------------------------------------*/ To see output from the commands go to the Stata results.Links in this page
Setting up a survey design and getting SEs Looking at the effect of alternative designs on precision Looking at regional rates Logistic regression Using replication methods Analyse SHlthS data in Stata Exemplar 3 ALL THE COMMANDS IN THIS SECTION CAN ALSO BE RUN FROM DIALOGUE BOXES back to top/*---------------------------------------------------- first set up the survey design and view its properties with svydes note that we need strata within regions (regstrat) ---------------------------------------------------------*/ svyset [pwei=weighta],psu(psu) strata(regstrat) set more off svydes /*------------------------------------------------------ you should find that you have strata each with 2 or (in a few cases) 3 PSUs Now get proportions in cigarette smoking categories and their standard errors ---------------------------------------------------------*/ svyprop cigst1 /*----------------------------------------------------------- svyprop does not give design effects or confidence intervals to get these for smokers you need to recode to a 0/1 variable and get its mean value -----------------------------------------------------------*/ recode cigst1 (-9 -8 -6 =.) (-1 1 2 3=0) (4=1),gen(smoker) svyprop smoker svymean smoker,deff deft ci /*---------------- check OK---------------------*/ table cigst1 smoker back to top
post-stratified , but to get the right SEs we need to redo the poststaritification and carry it out on each of the replicates --------------------------------------------------------------------------------------------------*/ collapse (sum) rtot=weight, by (region) /* make data file with totals*/ sort region save rtots,replace /* sort and save it*/ use ex3,clear sort region /* get original file and sort by region*/ merge region using rtots /*----------merge with totals and save as new file--------------*/ save ex3new,replace drop _merge /*---------- or next bit will fail------------------*/ tab ageg gen agesex=ageg*100+sex/* make an age sex varaible---*/ save ex3new,replace tab agesex collapse (sum) asext=weight, by (agesex) /* make data file with totals*/ sort agesex save astots,replace /* sort and save it*/ use ex3new,clear sort agesex /* get original file and sort by ctband*/ merge agesex using astots /*----------merge with totals and save--------------*/ save ex3new,replace /*----------------------------------------------- you now have a file with agesex and region totals -----------------------------------------------------*/ /*------------------- ------------------------------------------------------------------ and make a set of jacknife weights for this survey This next command will create 312 new variables (one for each replicate) where one of the 312 PSUs is dropped from each replication. Look at the data to check this ----------------------------------------------------------------------------------------------------------*/ survwgt create jkn, psu(psu) weight(weight) strata(regstrat) /*---------------- now use the survey replication commands--------------------------------*/ survwgt rake [all] , by(agesex region) totvars( asext rtot) replace save ex3reps,replace /*----------------------------------------------------------- recalculate smoker variable, as above -----------------------------------------------------------*/ recode cigst1 (-9 -8 -6 =.) (-1 1 2 3=0) (4=1),gen(smoker) /*----------------------------------------------- now use the command to get the mean and design effect for smokers using a jacknife method ------------------------------------------------------*/ svrmean smoker