The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » General » Comparison of Cluster Numbers over Several Surveys (Comparison of Cluster Numbers over Several Surveys)
Comparison of Cluster Numbers over Several Surveys [message #21841] Tue, 29 December 2020 05:16 Go to next message
Bianca is currently offline  Bianca
Messages: 5
Registered: December 2020
Member
Dear DHS-Community,

I am perfomring a large-scale data analysis combining several IR files from different countries being conducted between 2003 - 2018. Therefore, I need to interpret the cluster values (variable v001) in a right manner. In several publications I have noticed that the values of v001 for one country are combined. Thus, the total number of cluster is significantly lower than the total number of women being interviewed.

However, my understanding was that the cluster classification (e.g. v001 ="1") cannot be compared over different waves and that in each survey the clusters are classified uniquely and thus cannot be aggregated per country. Following my logic, the total number of cluster must equal the total number of women being interviewed. Is that right?

Thank you very much for your help in advance!

Warm greetings from Germany,
Bianca
Re: Comparison of Cluster Numbers over Several Surveys [message #21845 is a reply to message #21841] Tue, 29 December 2020 19:48 Go to previous messageGo to next message
Bridgette-DHS is currently offline  Bridgette-DHS
Messages: 2306
Registered: February 2013
Senior Member

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

Your logic is correct, the survey clusters cannot be combined across surveys; it is a survey related variable. Therefore when several surveys are combined across surveys, v001 should be reconstructed so that each cluster is unique throughout the combined dataset, for examples, when two surveys are combined with 100 clusters in survey 1 and 150 clusters in survey 2, the combined v001 variable should have 250 clusters. With that said, your sentence of "the total number of cluster must equal the total number of women being interviewed" is not correct. The total number of clusters has to be less than the number of completed women as we interview 20-30 HHs per cluster.

Re: Comparison of Cluster Numbers over Several Surveys [message #21849 is a reply to message #21845] Wed, 30 December 2020 05:27 Go to previous message
Bianca is currently offline  Bianca
Messages: 5
Registered: December 2020
Member
Thank you very much - Dr. Elkasabi and Bridgette! The reply was very straightforward and comprehensive. Have a good day!
Previous Topic: DHS Questionnaire questions --> Indicators on STATCompiler?
Next Topic: Missing variables in Albania DHS: SM901, SM903, S1201, S1203
Goto Forum:
  


Current Time: Thu Jun 17 19:12:50 Coordinated Universal Time 2021