The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Data » Weighting data » Calculating level weight for multicounty data - determining level of alpha to use
Calculating level weight for multicounty data - determining level of alpha to use Fri, 18 February 2022 10:08
 cw21 Messages: 4Registered: May 2021 Member
I am calculating the level weights for a multi-country (n=4) study using the DHS guide (https://www.dhsprogram.com/pubs/pdf/MR27/MR27.pdf) to calculation of level weight for a multilevel logistic regression. Level 1 = Individuals, level 2 = clusters and level 3 = country (no level weights are applied to the country levels as it is assumed that there is equal probability to the country being included and is therefore not down to the sampling procedure).

Firstly, I have undertaken the analysis within each country separately (with a two-level model) and determined the determined the which level of alpha should be used (e.g., allocation of variation in weights to the level-1 and level-2 units). As the level of alpha is different for each of the countries, I am unsure of how to deal with the alpha level when combining the data into a multi-country data set. Do I need to re-evaluate the alpha to be used in the combined dataset or could I use the appropriate level weights based on alpha for each country?

For example:

Country A - the most appropriate allocation of variation in weights is alpha = 0.25
Country B - the most appropriate allocation of variation in weights is alpha = 0.50
Country C - the most appropriate allocation of variation in weights is alpha = 0.75

Therefore, is it possible to use these level weights in the combined dataset, despite having a different level of alpha. Or is it best to use the same level of alpha throughout?

Many thanks in anticipation of your help
Re: Calculating level weight for multicounty data - determining level of alpha to use [message #24081 is a reply to message #24071] Mon, 21 February 2022 08:19
 Bridgette-DHS Messages: 2537Registered: February 2013 Senior Member

Following is a response from DHS Senior Sampling Specialist, Mahmoud Elkasabi:

Yes, level weights can be approximated separately by survey using different alphas.

Re: Calculating level weight for multicounty data - determining level of alpha to use [message #24082 is a reply to message #24081] Mon, 21 February 2022 08:23
 cw21 Messages: 4Registered: May 2021 Member
Thank you very much for your quick response.
 Previous Topic: Selecting sample within one standard deviation in R Next Topic: Adjusting weights on binary variables
Goto Forum:

Current Time: Sun Jul 3 10:41:29 Coordinated Universal Time 2022