Multistate Mark–Recapture Model Selection Using Score Tests

Authors

  • Rachel S. McCrea,

    Corresponding author
    1. National Centre for Statistical Ecology, School of Mathematics, Statistics and Actuarial Science, University of Kent, Canterbury CT2 7NF, U.K.
      email: R.S.McCrea@kent.ac.uk
    Search for more papers by this author
  • Byron J. T. Morgan

    1. National Centre for Statistical Ecology, School of Mathematics, Statistics and Actuarial Science, University of Kent, Canterbury CT2 7NF, U.K.
    Search for more papers by this author

email: R.S.McCrea@kent.ac.uk

Abstract

Summary Although multistate mark–recapture models are recognized as important, they lack a simple model-selection procedure. This article proposes and evaluates a step-up approach to select appropriate models for multistate mark–recapture data using score tests. Only models supported by the data require fitting, so that over-complicated model structures with too many parameters do not need to be considered. Typically only a small number of models are fitted, and the procedure is also able to identify parameter-redundant and near-redundant models. The good performance of the technique is demonstrated using simulation, and the approach is illustrated on a three-region Canada goose data set. In this case, it identifies a new model that is much simpler than the best model previously considered for this application.

Ancillary