This package provides tools to implement and evaluate designs to measure the prevalence of hidden populations.
Researchers use a variety of techniques that vary on three dimensions: how and from whom data is gathered, which measures are taken, how analysis is done using these measure. Specific measures range from simple population sample with scaling by inverse sampling weights, to complex sampling designs using respondent driven sampling or time location sampling, coupled with network scale-up methods, or Bayesian analyses. Each method involves many specific design choices, and the relative performance of different methods likely varies over applications.
The tools provided here help researchers select designs, assess their quality, and implement their analyses. In addition the tools can allow researchers to pool findings from multiple methods in order to implement meta-analysis that lets researchers learn about the relative effectiveness of approaches.
We rely on
for evaluating designs via simulation. The key idea of the simulation
workflow is that we simulate the design as a set of steps. When
concatenated the steps produce an instance of a design. When replicated
these instances let us assess the overall performance of a design.
Fully specified designs include information on:
Each of these MIDA steps can be defined using “handlers” from the
hiddenmeta package and can be adjusted to reflect details
of individual study designs.
The individual study design simulation involves:
Simulated study level results feed into the meta-analysis:
Conduct multi-study design for as many sampling-estimator pairs in each study as possible, then diagnose the multi-study design. Calculate average (across simulations) estimand and bias of sampling-estimator for each of the studies and estimator sampling strategies. These will serve as population we will be drawing population-sampling-estimator triads
Sampling consists of drawing population-sampling-estimator triads presuming that each study uses at least two sampling strategies at a time
Once we draw sample we use Stan model to estimate study-specific estimands and sampling-estimator specific errors (biases, study level prevalence or size, cost-effectiveness)
hiddenmetaworkflow please read this vignette
This project is based at the WZB Berlin Social Science Center (IPI group) and is generously supported by a grant from the African Programming & Research Initiative to End Slavery (APRIES).