Abstract:
|
Increasingly, surveys have to rely on more than one sample source to improve coverage and/or secure the needed sample size in a defensible and cost-effective manner. As such, two or more independent samples are selected from separate sampling frames with varying representations of the target population of interest. In many instances this involves combining probability and nonprobability samples, particularly when nonprobability samples from online panels are used to augment a main probability-based sample. This paper provides an overview of weighting refinements that can improve the external validity of data from such surveys, including a method for optimal integration of surveys.
|