Adaptive Sampling Designs and Associated Estimators
Open Access
- Author:
- Dryver, Arthur Lance
- Graduate Program:
- Statistics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 11, 1999
- Committee Members:
- Bing Li, Committee Member
Linda Marie Collins, Committee Member
Jogesh Babu, Committee Member
Steve K Thompson, Committee Chair/Co-Chair - Keywords:
- adaptive sampling
sampling designs - Abstract:
- In conventional sampling, we generally use different estimators depending on how the sample was obtained. If a simple random sample was taken the sample mean would be used, assuming the parameter of interest is the population mean. On the other hand, if a sample was taken with probability proportional to size, we would use the Horvitz-Thompson estimator (Horvitz and Thompson 1952). In that case, the sample mean would yield a biased estimate. In adaptive sampling, which estimator is most appropriate to use also depends on how the sample was obtained. This dissertation will cover previous work in adaptive sampling when the initial sample is taken without replacement of units and the case when the sample is taken without replacement of networks. It will also cover new estimators when the initial sample is taken without replacement of units and on adaptive cluster sampling without replacement of clusters. The latter estimators proposed are design unbiased estimators. Often, a sample cannot be taken in the manner necessary to utilize design unbiased estimators and for this reason model based estimators are important to develop. Maximum likelihood model based estimators for estimating population size utilizing adaptive snowball sampling will be covered.