Use of the Barker model in an experiment examining covariate effects on first-year survival in Ross's Geese ( Chen rossii ): A case study

Abstract

The Barker model provides researchers with an opportunity to use three types of data for mark-recapture analyses - recaptures, recoveries, and resightings. This model structure maximizes use of encounter data and increases the precision of parameter estimates, provided the researcher has large amounts of resighting data. However, to our knowledge, this model has not been used for any published ringing studies. Our objective here is to report our use of the Barker model in covariate-dependent analyses that we conducted in Program MARK. In particular, we wanted to describe our experimental study design and discuss our analytical approach plus some logistical constraints we encountered while conducting a study of the effects of growth and parasites on survival of juvenile Ross's Geese. Birds were marked just before fledging, alternately injected with antiparasite drugs or a control, and then were re-encountered during migration and breeding in following years. Although the Barker model estimates seven parameters, our objectives focused on annual survival only, thus we considered all other parameters as nuisance terms. Therefore, we simplified our model structures by maintaining biological complexity on survival, while retaining a very basic structure on nuisance parameters. These analyses were conducted in a two-step approach where we used the most parsimonious model from nuisance parameter analyses as our starting model for analyses of covariate effects. This analytical approach also allowed us to minimize the long CPU times associated with the use of covariates in earlier versions of Program MARK. Resightings made up about 80% of our encounter history data, and simulations demonstrated that precision and bias of parameter estimates were minimally affected by this distribution. Overall, the main source of bias was that smaller goslings were too small to retain neckbands, yet were the birds that we predicted would have the lowest survival probability and highest probability for parasite effects. Consequently, we considered our results conservative. The largest constraint of our study design was the inability to partition survival into biologically meaningful periods to provide insight into the timing and mechanisms of mortality.