Non-Probability Samples from Populations That Are Identified Through Costly Screening

• Activity Number: 304
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Survey Research Methods Section
• Author(s): K. P. Srinath*+ and Charles DiSogra
• Companies: Abt SRBI and Abt SRBI
• Keywords: precision ; bias ; screening cost

Abstract

In recent years, non-probability samples are being selected in more and more surveys. This may be driven by lack of an accessible frame or non-existence of a suitable frame for probability sampling. More often cost and practicality are factors. When the eligible population is rare, it has to be identified through screening a general population. In such surveys, the use of probability samples can be very expensive because of high screening costs to locate eligible cases. Low response rates may also make the results unreliable because of bias due to nonresponse and high variance. We examine the use of non-probability samples in such surveys. A major concern about the results from non-probability samples is the unknown bias in the estimates. We consider and present steps that could be taken before and after selection of a non-probability sample to minimize bias. We also consider methods of making the non-probability sample close to one of the possible samples that a probability design could produce. We investigate the possibility of assigning a measure of precision to the estimates from non-probability samples and discuss the pros and cons of such a metric.