Estimating transitions between states using measurements with imperfect detection: application to serological data

Classifying the states of an individual and quantifying transitions between states are crucial while modeling animal behavior, movement, and physiologic status. When these states are hidden or imperfectly known, it is particularly convenient to relate them to appropriate quantitative measurements taken on the individual. This task is, however, challenging when quantitative measurements are not available at each sampling occasion. For capture-recapture data, various ways of incorporating such non-discrete information have been used, but they are either ad hoc and/or use a fraction of the available information by relying on a priori thresholds to assign individual states. Here we propose assigning discrete states based on a continuous measurement, and then modeled survival and transition probabilities based on these assignments. The main advantage of this new approach is that a more informative use of the non-discrete information is done. As an illustrative working example, we applied this approach to eco-epidemiological data collected across a series of years in which individuals of a long-lived seabird, the Black-legged Kittiwake (Rissa tridactyla), could either be visually detected or physically recaptured and blood sampled for subsequent immunological analyses. We discuss how this approach opens many perspectives in eco-epidemiology, but also more broadly, in population ecology.