The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » General » svyset-Multilevel modeling (Some Information is missing in final report of PDHS-17 for calculating weights)
svyset-Multilevel modeling [message #19913] Fri, 28 August 2020 16:50 Go to next message
Rehman is currently offline  Rehman
Messages: 3
Registered: August 2020
Member
Hi DHS team,

We are following your recent guidelines in 'MULTILEVEL MODELING USING DHS SURVEYS: A FRAMEWORK TO APPROXIMATE LEVEL-WEIGHTS' file for weights calculation.
In the sample file, your team has used the zimbabwe final report. Some of the information like the average size in the number of households (stage 1 step 3) is not mentioned in the PDHS-17 final report (table A.1). There is only 'Number of enumeration blocks by region' is available which is not enough to calculate weights.

Any help in that regard would be highly appreciated.

Thank you!
Re: svyset-Multilevel modeling [message #19940 is a reply to message #19913] Fri, 04 September 2020 10:10 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Following is a response from DHS Senior Sampling Specialist, Mahmoud Elkasabi:

It is true that the average size of EAs was not provided in the final report of the 2017-18 Pakistan DHS. Fortunately, such information is available, even on the block level, on the website of Pakistan Bureau of Statistics. You will need to compile the district-level numbers of households to produce the number of households by state. This distribution along with the distribution of the EBs in A.1 can be used to produce the average size per each state.

[Updated on: Fri, 04 September 2020 10:31]

Report message to a moderator

Re: svyset-Multilevel modeling [message #19946 is a reply to message #19940] Sat, 05 September 2020 05:16 Go to previous messageGo to next message
Rehman is currently offline  Rehman
Messages: 3
Registered: August 2020
Member
Hi,

Thank you so much. The information you provided is really helpful.
However we have two follow up questions:

1: The similar information is not available for 2012-13 dataset even on the website. The information we found was total population based on 1998 population census and not the number of households. Is there a way out?

2: The guideline of weights are for two level, one is cluster and other is women/men/household. We are using three level modelling where we also wanted to add third level (children) as one mother has more than one child for most cases. Form reading the sampling procedure of DHS, it can be assume that children have equal probability to be in sample, so can we add weight=1 for third level?

Regards,
Re: svyset-Multilevel modeling [message #19962 is a reply to message #19946] Tue, 08 September 2020 15:49 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 3016
Registered: February 2013
Senior Member
Following is another response from DHS Senior Sampling Specialist, Mahmoud Elkasabi:

1. For 2012-13 dataset, you can use the population distribution from the 1998 census. Usually distributions of households and population are good proxies for each other's.

2. We do not have specific guidelines for the 3-level case. I believe using weight = 1 for the third level should be OK.
Re: svyset-Multilevel modeling [message #19972 is a reply to message #19962] Wed, 09 September 2020 03:12 Go to previous message
Rehman is currently offline  Rehman
Messages: 3
Registered: August 2020
Member

We really appreciate your prompt response. Thank you!

Cheers,
Previous Topic: Replacements?
Next Topic: Imputing missing DOBs for children
Goto Forum:
  


Current Time: Thu Mar 28 11:41:04 Coordinated Universal Time 2024