The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Wealth Index » Calculating a Representative Wealth Index for Clusters Using DHS Sample Weights
Re: Calculating a Representative Wealth Index for Clusters Using DHS Sample Weights [message #30669 is a reply to message #30657] Fri, 17 January 2025 05:30 Go to previous messageGo to previous message
Rean is currently offline  Rean
Messages: 3
Registered: January 2025
Member
Thank you so much for your insightful response!

Regarding the use of weighted and unweighted means for representing the cluster-level Wealth Index (WI), I would like to confirm my understanding. Since all households within a cluster share the same weight, I am using two approaches to derive a representative value for the cluster's wealth level for machine learning training labels. The two methods are:

1. Unweighted mean WI: The average WI for all households within the cluster.
2. Weighted mean WI: The average WI of all households within the cluster, weighted by the cluster's associated weight (hv005).

In my machine learning experiments, I observed that the model's performance was slightly better when using the weighted WI as training labels rather than the unweighted mean WI. Could the analysis you mentioned, which involves weighting by the cluster's weight, explain this result? In other words, using the weighted mean WI as the label for the clusters' overall wealth status might provide a more representative and useful feature for training the model.

I greatly appreciate your time and valuable insights, which have truly helped me a lot regarding this issue!
 
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: inside the state wealth index
Goto Forum:
  


Current Time: Mon Feb 3 00:02:14 Coordinated Universal Time 2025