I am working with survey data that contain information at the individual, household and community level. The survey is representative at the national and first-administrative level but provides sampling weights only for individuals and households.
The sampling strategy consists in a two-stage cluster design with the following characteristics:
- stratification by first-administrative level;
- optimum allocation distribution for first-administrative level (which minimizes the standard error);
- selection with probability proportional to the number of households (PPS).
Moreover, within each of first-administrative unit (strata) EAs were selected as PSUs in the first sampling stage (based on PPS). In the second stage a cluster of ten households was selected from the updated household listing, based on a fixed-interval and random-start procedure.
The second level of sample clustering represents a community.
I am willing to run analyses at the community level but the corresponding sampling weights are not provided. How could I weight the communities such that my estimates will be representative at the national level?