Estimating Change Over Time From Aggregate Samples Plus Partial Transition Data
Abstract
**Please note that the full text is embargoed** ABSTRACT: As such, the paper has two basic thrusts. The first is that to obtain efficient estimates of change measures (e.g. of transition probabilities) based on aggregate data, auxiliary data will in general be needed; several potentially useful auxiliary sampling schemes are discussed. The second thrust is the manner in which estimates from the such varied sampling schemes can be combined to achieve such efficient estimates. Since it either directly or indirectly suggests many possible longitudinal study designs, the paper necessarily cannot provide real examples of all (or even most) of the study designs considered. Indeed, the main point here is to suggest how future longitudinal studies aimed at measuring change, but in circumstances where it is difficult to obtain transition data, might be done more effectively.