Estimation of bird distribution based on ring re-encounters: precision and bias of the division coefficient and its relation to multi-state models

TY - JOUR

T1 - Estimation of bird distribution based on ring re-encounters

T2 - precision and bias of the division coefficient and its relation to multi-state models

AU - Korner-Nievergelt, Fränzi

AU - Schaub, Michael

AU - Thorup, Kasper

AU - Vock, Michael

AU - Kania, Wojciech

PY - 2010/2

Y1 - 2010/2

N2 - Capsule: The division coefficient is an estimate of the proportion of ringed birds migrating to different destination areas taking into account area-specific re-encounter probabilities. Aims: To explore precision and bias of the division coefficient method by a simulation study and to compare the approach with multi-state models. Methods: In a simulation study true and estimated division coefficients were compared. The division coefficient method was mathematically compared with the multi-state model. Results: The estimated division coefficients seemed to be unbiased if the assumptions were met. The precision decreased if the bird distribution became similar in both bird groups and when difference between area-specific re-encounter probabilities increased. A bootstrap method to assess precision is presented. The estimates from the division coefficient method equal the maximum likelihood estimates in a multi-state model including only one time interval. Conclusion: Before applying the division coefficient method or a multi-state model to real data a simulation study should be conducted in order to explore the behaviour of parameter estimation. The division coefficient method with the bootstrap confidence intervals is an easy alternative to a multi-state model with one time interval when the bird distribution between destination areas (e.g. migratory connectivity) alone is of interest.

AB - Capsule: The division coefficient is an estimate of the proportion of ringed birds migrating to different destination areas taking into account area-specific re-encounter probabilities. Aims: To explore precision and bias of the division coefficient method by a simulation study and to compare the approach with multi-state models. Methods: In a simulation study true and estimated division coefficients were compared. The division coefficient method was mathematically compared with the multi-state model. Results: The estimated division coefficients seemed to be unbiased if the assumptions were met. The precision decreased if the bird distribution became similar in both bird groups and when difference between area-specific re-encounter probabilities increased. A bootstrap method to assess precision is presented. The estimates from the division coefficient method equal the maximum likelihood estimates in a multi-state model including only one time interval. Conclusion: Before applying the division coefficient method or a multi-state model to real data a simulation study should be conducted in order to explore the behaviour of parameter estimation. The division coefficient method with the bootstrap confidence intervals is an easy alternative to a multi-state model with one time interval when the bird distribution between destination areas (e.g. migratory connectivity) alone is of interest.

U2 - 10.1080/00063650903154439

DO - 10.1080/00063650903154439

M3 - Journal article

SN - 0006-3657

VL - 57

SP - 56

EP - 68

JO - Bird Study

JF - Bird Study

IS - 1

ER -