|
Abstract:
|
Repeated surveys of a population over time can provide reliable estimates of many population quantities, as well as on changes over time. However, there is often a significant delay between a survey being administered and the release of a final data product. Other, less statistically robust data sources are often available more quickly. Incorporating non-survey auxiliary information into a model may allow for providing more timely and more accurate predictions and estimates. We develop a unit-level Markov model for categorical data, which incorporates survey observations and auxiliary information. The model provides simple, closed-form estimates of transition probabilities between categories for each unit, and incorporates more recent auxiliary information with the survey data. Survey weights can be naturally incorporated to provide estimates of population totals. We demonstrate the model on data taken from the Natural Resources Inventory, using the Cropland Data Layer for auxiliary unit level information.
|
