I'm analyzing some country-level data that was sampled in the following way:
1) Country divided in five different Zones - North, South, Centre, East, West;
Within each zone, 1-3 Regions were selected (each zone had between 2-4 regions)
***here I have my first problem: "Note: Missing standard error because of stratum with single sampling unit."
2) In the second stage, 2-4 Districts were selected within each chosen region (each region had between 3-10 districts)
3) In the third stage, 3-6 Facilities were selected within each chosen district (each district had between 6-32 facilities)
I have therefore different probabilities of selecting facilities from different districts, regions, zones. I calculated that probability for the different levels:
Code:
p_selection_region=n_sampled_regions/n_regions p_selection_district=p_selection_region*(n_sampled_districts/n_districts) p_selection_facility=p_selection_region*p_selection_district*(n_sampled_facilities/n_facilities)
I then tried to define my survey design, but I am not sure how to do it. Based on "Example 2: Multistage survey data" from the STATA SURVEY DATA REFERENCE MANUAL I defined it like this:
Code:
svyset REGION [pw=weight_facility], strata(ZONE) fpc(n_regions) || DISTRICT, fpc(n_districts) || facility, fpc(n_facilities)
Code:
svyset REGION, weight(weight_region) strata(ZONE) fpc(n_regions) || DISTRICT, weight(weight_district) fpc(n_districts) || facility, weight(weight_facility) fpc(n_facilities)
Code:
svy: meglm y x || region: || district: || facility:, family(binomial) link(logit) or
Suggestions? What am I doing wrong?
Thank you!
0 Response to doubt in svyset for a 3-stage sampling
Post a Comment