YSPH Biostatistics Seminar: "Measures of Selection Bias for Proportions Estimated from Non-Probability Samples"
NOTE: BIS 525 students are required to attend in person (47 College St., Room 106A). All others are requested to attend via Zoom.
SPEAKER: Rebecca Andridge, PhD, Associate Professor, Department of Biostatistics, The Ohio State University
TITLE: “Measures of Selection Bias for Proportions Estimated from Non-Probability Samples”
ABSTRACT: The proportion of individuals in a finite target population that has some characteristic of interest is arguably the most commonly estimated descriptive parameter in survey research. Unfortunately, the modern survey research environment has made it quite difficult to design and maintain probability samples: the costs of survey data collection are rising, and high rates of nonresponse threaten the basic statistical assumptions about probability sampling that enable design-based inferential approaches. As a result, researchers are more often turning to non-probability samples to make descriptive statements about populations. Non-probability samples do not afford researchers the protection against selection bias that comes from the ignorable sample selection mechanism introduced by probability sampling, and descriptive estimates based on non-probability samples may be severely biased as a result. In this seminar I describe a simple model-based index of the potential selection bias in estimates of population proportions due to non-ignorable selection mechanisms. The index depends on an inestimable parameter that captures the amount of deviation from selection at random; this parameter ranges from 0 to 1 and naturally lends itself to a sensitivity analysis. I describe both maximum likelihood and Bayesian approaches to estimating the index, and illustrate its use via simulation and via application to real data.
Speaker
The Ohio State University
Rebecca Andridge, PhDAssociate Professor