Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Steve Samuels <sjsamuels@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: Pooling DHS surveys: svyset command? |
Date | Wed, 25 Jul 2012 09:49:15 -0400 |
You would need to check whether any strata by either definition (v024x v025, v023xv025) have only one PSU (first-stage cluster). Such PSUs are called "certainty units" and there are several options (and opinions) on how to deal with them. If the v023xv025 definition avoids them, all the better. Steve sjsamuels@gmail.com On Jul 25, 2012, at 9:33 AM, <M.Vandemoortele@lse.ac.uk> <M.Vandemoortele@lse.ac.uk> wrote: Dear Steve, Thank you for your response. You are correct - the v024 variable does not designate regions in the Malawi 2000 DHS. Thankfully a statistician who works for MeasureDHS has kindly clarified the issue: "I guess it was stratified by urban/rural crossing district defined by a country specific variable S006. But when you check the frequency, some urban stratum might have only one cluster selected. If this poses problems for your application, I suggest you to use V023 crossing V025, this gives you the best approximation of the stratification. Hope this helps." Very helpful. Best, Milo -----Original Message----- From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Steve Samuels Sent: 24 July 2012 15:38 To: statalist@hsphsun2.harvard.edu Subject: Re: st: Pooling DHS surveys: svyset command? Milo- If the information you provide is correct. our -svyset- statement is OK. However your description of the 2000 sample appears to be inaccurate, as 11 districts were designated for oversampling (www.measuredhs.com/pubs/pdf/FR123/FR123.pdf). These do not appear to be "regions", but rather sub-strata of the regions to which they belong. So I doubt that "v024" designates regions in 2000, but leave it to you to check. Steve sjsamuels@gmail.com On Jul 24, 2012, at 9:13 AM, <M.Vandemoortele@lse.ac.uk> <M.Vandemoortele@lse.ac.uk> wrote: Dear Stata list serve members, I am pooling three years of Malawi DHS data (2000, 2004 and 2010). How would you recommend that I take into account the survey design given psu and strata variables and adjustment weights? The sampling designs differ across surveys and seem to be as follows: - DHS 2010: stratified by district then urban/rural. Clusters were selected using probability-proportionate to size (PPS) (frame was the 2008 census enumeration units) - DHS 2004: stratified by region then urban/rural. Clusters were then clusters selected using PPS (Frame 1998 census enumeration units) - DHS 2000: stratified by region then urban/rural. Clusters were selected through systematic sampling (Frame 1998 census enumeration units) Instructions on the DHS website say do the following: - to generate weight: generate weight = v005/1000000 - to make unique strata values depending on how sampling design (in this case already done for 2010, and 2004 and 2000 v025 and v024 represent region and urban/rural variables): egen strata = group(v024 v025), label DHS website, however, does not appear to indicate how to take into account survey design when surveys are pooled. On the Stata ListServe the following recommendations are provided on a related, but non-DHS, question (http://www.stata.com/statalist/archive/2008-10/msg00521.html): If your surveys were stratified, to begin with, then it would become: svyset psuXyear [pw=weight in each wave], strata(waveXoriginal_strata) where -X- stands for interaction along the lines of: egen psuXyear = group(psu year) Svyset command I use is as follows: svyset psuXyear [pw=weight], strata(strataXyear) singleunit(scaled) For pooling DHS surveys, does this svyset command look appropriate? Best regards, Milo Please access the attached hyperlink for an important electronic communications disclaimer: http://lse.ac.uk/emailDisclaimer * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ Please access the attached hyperlink for an important electronic communications disclaimer: http://lse.ac.uk/emailDisclaimer * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/