|
Abstract:
|
The increasing use of internet panel surveys has led to a growing need for inferences from these samples that are generally selected with non-probability sampling methods from a very large panel population. Also lacking is evidence of the validity of these samples at least in the sense of representation, coverage, and comparability with probability samples of similar populations. This paper first describes the weighting methods used to generate valid estimates. It then explores the development of variances and confidence intervals for these weighted samples, and uses these to compare the survey estimates with probability sample estimates. This research uses data from a series of internet panel studies; specifically, we used the data from the study conducted in Los Angeles County. We compare the results with local data from the Behavioral Risk Factor Surveillance System (BRFSS) and from the California Health Interview Survey (HIS).
|
