Abstract:
|
One factor that is often overlooked in dynamic adaptive designs is whether a certain data collection strategy is "worth it", with respect to reaching data collection goals. Ideally, we would not send that expensive field interviewer to attempt a case that will not respond to either a telephone attempt or a personal visit, as resources could be better spent on other similar cases. Alternatively, perhaps a small incentive would convince a sample individual to respond. To use this type of information, we need to look at the response propensity of a case (and how it is affected by different data collection features) in addition to its value to the data collection goal. This is difficult, as we do not know in advance the effect of different combinations of features on survey response - we cannot tell the future. As a result we rely on models, but early in data collection, when a decision could have a large impact at reducing cost and burden, response propensities are often unreliable, shown by Wagner. This talk will illustrate how poor estimates of response propensity can affect data collection decisions, the potential benefits of external information, and how to integrate that data.
|