Chapter 4 Errors of nonobservation: Sampling and coverage

4.1 Coverage of the population of interest

We discuss selection bias, that is, systematic error that stems from the fact that people who own a specific device or use a technology - and thus are able to provide data - might differ in the variables of interest from those who don’t.

4.2 Opt-in and Probability Samples

We discuss the benefits and drawbacks of these two sample types. Probability samples are more expensive but provide a statistical basis for inference. Opt-in samples may be biased if those who opt-in are different than those who do not.

4.3 Device distribution considerations

Some studies rely on participants to provide their own smartphone or device, and others distribute devices to participants. This choice again affects selection bias in the collected data.

4.4 Longitudinal data collection

Given the cost and difficulty of recruiting participants to a WAS study. It may make sense to keep participants involved for weeks or months. This approach switches the focus from point in time estimates to longitudinal estimates.

4.5 References