Design variations in adaptive web sampling

Kyle Vincent successfully defended his M.Sc. project entitled "Design variations in adaptive web sampling" on 25 July 2008.

There is an increasing body of literature related to sampling for network and spatial settings. Although current link-tracing methods like adaptive cluster sampling, snowball sampling, and targeted random walk designs have advantages over conventional designs, some of the following drawbacks remain evident: there is a lack of flexibility in sample placement; there is an inability to control over sample sizes; and efficiency gains over conventional sampling designs for estimating population parameters may not be achievable. Adaptive web sampling (AWS) is a recently developed link-tracing design that overcomes some of these issues. Furthermore, the flexibility inherent to the AWS method permits many design variations. Using a simulated network population, an empirical population at risk for HIV/AIDS, a simulated spatial population, and an empirical population of birds, this project performs a simulation study to compare the performance of three variations of AWS strategies.

Keywords: Adaptive sampling, Link-tracing designs, Markov chain Monte Carlo, Network sampling, Rao-Blackwellization, Spatial sampling

This type of interdisciplinary work is a hallmark of our program in Applied Statistics at Simon Fraser University. For more information, please contact Kyle Vincent (kvincent@sfu.ca) or his supervisor Steve Thompson (thompson@sfu.ca), Department of Statistics and Actuarial Science.

2008-07-27