EVALUATING PROPAGATION METHOD PERFORMANCE OVER TIME WITH BAYESIAN UPDATING: AN APPLICATION TO INCUBATOR TESTING

Authors

SARAH J. CONVERSE, USGS Patuxent Wildlife Research CenterFollow
JANE N. CHANDLER, USGS Patuxent Wildlife Research Center
GLENN H. OLSEN, USGS Patuxent Wildlife Research Center
CHARLES C. SHAFER, USGS Patuxent Wildlife Research Center

Date of this Version

2010

Document Type

Article

Citation

Converse, S.J., J.N. Chandler, G.H. Olsen, and C.C. Shafer. Evaluating propagation method performance over time with bayesian updating: an application to incubator testing. In: Hartup, Barry K., ed., Proceedings of the Eleventh North American Crane Workshop, Sep 23-27, 2008, Wisconsin Dells, Wisconsin (Baraboo, WI: North American Crane Working Group, 2010), pp. 110-117.

Comments

Reproduced by permission of the North American Crane Working Group.

Abstract

In captive-rearing programs, small sample sizes can limit the quality of information on performance of propagation methods. Bayesian updating can be used to increase information on method performance over time. We demonstrate an application to incubator testing at USGS Patuxent Wildlife Research Center. A new type of incubator was purchased for use in the whooping crane (Grus americana) propagation program, which produces birds for release. We tested the new incubator for reliability, using sandhill crane (Grus canadensis) eggs as surrogates. We determined that the new incubator should result in hatching rates no more than 5% lower than the available incubators, with 95% confidence, before it would be used to incubate whooping crane eggs. In 2007, 5 healthy chicks hatched from 12 eggs in the new incubator, and 2 hatched from 5 in an available incubator, for a median posterior difference of <1%, but with a large 95% credible interval (-41%, 43%). In 2008, we implemented a double-blind evaluation method, where a veterinarian determined whether eggs produced chicks that, at hatching, had no apparent health problems that would impede future release. We used the 2007 estimates as priors in the 2008 analysis. In 2008, 7 normal chicks hatched from 15 eggs in the new incubator, and 11 hatched from 15 in an available incubator, for a median posterior difference of 19%, with 95% credible interval (-8%, 44%). The increased sample size has increased our understanding of incubator performance. While additional data will be collected, at this time the new incubator does not appear adequate for use with whooping crane eggs.