The DHS Program User Forum: Sampling

Home » Data » Sampling » Stratification

Show: Today's Messages :: Show Polls :: Message Navigator

Re: Stratification [message #1475 is a reply to message #1474]

Mon, 03 March 2014 14:49

Trevor-DHS
Messages: 805
Registered: January 2013

Senior Member

I can't find anything that I can provide that really gives justification for this. The approach was used throughout the World Fertility Surveys (WFS) with systematic sampling, and was continued for a considerable time in DHS. You can find reference to implicit stratification in http://unstats.un.org/unsd/demographic/meetings/egm/Sampling _1203/docs/no_2.pdf. you may also find more if you search for "implicit stratification in sampling" on Google.

The documentation for the "Clusters" software used by the WFS to calculate sampling errors contains the following statements:
"It should be noted that the strata used for computation of sampling errors are not necessarily identical to the original explicit strata used in the sample selection. The difference between the two may arise for two main reasons:
1) Whenever PSUs are selected by systematic sampling* from an ordered list, adjacent units should be paired or grouped to form new smaller strata which are used for sampling error computations.
2) Sampling error computations require that there be at least 2 PSUs per stratum. Any strata for which only one PSU has been selected must be 'collapsed' together to form pairs (or other groups) of PUSs. These constitute new strata to be used for sampling error computations. Such grouping is done on the basis of characteristics of the whole strata population (pairing most similar strata), and not on the characteristics of the selected PSUs. Collapsing of the strata in this way leads to slight over estimation of the sampling error.
* 'systematic sampling' means selection, at a fixed interval, from a list starting from a randomly determined point."

This doesn't provide a justification, and unfortunately I only have a very old paper version of the documentation (that is falling apart), but does describe the procedure that was used in WFS and the early part of DHS.

Report message to a moderator

[Message index]

		Stratification By: hlyons on Mon, 03 February 2014 20:09
		Re: Stratification By: Bridgette-DHS on Wed, 26 February 2014 09:20
		Re: Stratification By: hlyons on Thu, 27 February 2014 14:52
		Re: Stratification By: Trevor-DHS on Fri, 28 February 2014 11:40
		Re: Stratification By: hlyons on Fri, 28 February 2014 14:35
		Re: Stratification By: Trevor-DHS on Mon, 03 March 2014 12:12
		Re: Stratification By: hlyons on Mon, 03 March 2014 13:12
		Re: Stratification By: Trevor-DHS on Mon, 03 March 2014 14:49
		Re: Stratification By: Trevor-DHS on Mon, 03 March 2014 15:13
		Re: Stratification By: hlyons on Mon, 03 March 2014 18:08

Previous Topic:	Specific description of the data
Next Topic:	Two or Three Sampling Stages

Goto Forum:

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Thu Aug 7 05:19:32 Coordinated Universal Time 2025