Abstract

Background
The nonresponse weighting adjustment of the PHIA surveys employs the weighting class method in combination with a tree analysis to identify predictors significant to response propensity. Variable selection for this type of nonresponse adjustment identifies auxiliary variables correlated with response propensity alone and produces one set of weights applicable for all analyses of the survey data. An alternative approach identifies auxiliary variables correlated to both the response probability and selected key outcome variables. This approach may identify a different set of variables for the nonresponse adjustments and may produce more efficient estimates for the key outcome variables.

Setting
The Population-based HIV Impact Assessment (PHIA) surveys from 2016-2017

Methods
Weighting class, joint-classification, and two-step modeling

Results
There was little difference among estimates produced by the alternative weighting methods and the PHIA estimates. The joint-classification method produced more efficient estimates (i.e., smaller design effects) compared to the PHIA method, while the two-step method was inconclusive.

Conclusions
The efficiency of the estimates produced by the PHIA weighting method closely resembles those specifically targeted at key survey outcomes and serves well as a multi-purpose weight.