The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Continuous Survey Weights
Continuous Survey Weights [message #2335] Wed, 04 June 2014 09:39 Go to next message
Laura_Nolan is currently offline  Laura_Nolan
Messages: 1
Registered: June 2014
Member

Is there anything different about the continuous survey in terms of its representativeness (i.e. analysts' use of the survey weights)? An article about the use of DHS in the WHO Bulletin mentions "unique challenges involved in the statistical analysis of continuous survey data":

http://www.who.int/bulletin/volumes/90/8/11-095513/en/

What is this referring to? Should continuous survey weights be treated/used any differently than regular DHS survey weights?

Many thanks.
Re: Continuous Survey Weights [message #2336 is a reply to message #2335] Wed, 04 June 2014 09:57 Go to previous message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3017
Registered: February 2013
Senior Member
Following is a response from Senior DHS Specialist, Shea Rutstein:

The Peru Continuous DHS started its first cycle in 2004. From 2004 to 2007, the yearly sample was a 1/5th random sample of the clusters drawn for the 2000 Peru DHS, which was based 1999 pre-census sampling frame (the census never was realized. This sub-selection was done to reduce sampling errors for trends. Thus the principal component of the sample weights were the selection probabilities for those clusters in 2000. Adjustments were made to the weights for three items: the change in the number of dwellings between 2000 and the cycle year (a relisting of each cluster was performed before interviewing fieldwork); non-response rates for households and women; and post-stratification for better representation (given that it was a random subsample of a sample). In 2008, an additional large set of clusters was added to increase sample size. A different frame was used for the additional clusters and weights were calculated for the combined frames to represent urban and rural areas of each department (state) according to the census of 2007, as well as adjusting for non-response. Given that none of the clusters was reused, the sample weights can be used for individual years. From 2009 onward, the sample frame is based on the 2007 census, so that the sampling weights can also be used for comparisons without problems, even though 1/2 of the clusters of each year's sample was reused in the following year to reduce the sampling errors for trends.

Anthropometric measurements were taken in 2005, 2007, and from 2008 onward. From 2008, the survey recorded the height and weight of children under five years that was noted on their vaccination cards during Growth and Development Controls. These data include the date of the control and can serve as real cohort information. You may want to use them instead.

[Updated on: Wed, 04 June 2014 09:58]

Report message to a moderator

Previous Topic: When/How to use (de-normalized) weights for pooled data analysis (four waves of BD child anthro data
Next Topic: Pooled datasets - weighting data problem
Goto Forum:
  


Current Time: Fri Mar 29 03:15:40 Coordinated Universal Time 2024