Advisor(s) - Committee Chair
Reagan Brown, Xiaowen Chen, Katrina Burch
Department of Psychological Sciences
Master of Science
Nonprobability samples are often used in place of probability samples because the former are less trouble and less expensive. Unfortunately, it is difficult to determine how well a sample represents population parameters when using nonprobability samples. Researchers attempt to mitigate the disadvantages of nonprobability sampling by performing post hoc corrections, but this adjustment may not successfully undo the effects of nonprobability sampling. To examine these effects, a Monte Carlo simulation was conducted to create a pseudo-population from which samples were drawn. Forty-one conditions were replicated 10,000 times each, with each sample consisting of 100 observations. A post-stratification adjustment was made to these sample means as a post hoc correction. Confidence intervals were computed from the sample means before and after the adjustment. It was found that even slight correlations between the dependent variable and the likelihood of being sampled resulted in biased sample means and ineffective confidence intervals. Furthermore, post-stratification adjustments had mixed results but were generally ineffective in correcting for bias.
Applied Statistics | Industrial and Organizational Psychology | Physical Sciences and Mathematics | Psychology | Social and Behavioral Sciences | Statistical Methodology | Statistics and Probability
Hong, Julia, "A Monte Carlo Analysis of Nonprobability Sampling & Post Hoc Corrections" (2023). Masters Theses & Specialist Projects. Paper 3638.