How to incorporate resighting error? Input requested

[b06jll]

For the last two years I have been collecting data on the probability of incorrectly resighting bird bands and have come up with an estimate of the error.
Since this error obviously affects our ability to accurately quantify survival and detect declining populations, I would like to figure out a way to incorporate this error into survival analyses...perhaps through developing a new "band error" parameter or through some other measure.
I would really appreciate any and all feedback.
Thank you in advance!
Jennifer Lavers

[ganghis]

Jennifer,
The answer to your question depends on what types of errors are made. For instance, if the probability of mis-IDing an individual as being another (real) individual is low, you end up with too many histories being seen only once (e.g. the number of used codes is small enough that you will almost always end up with a new code if you mis-ID an animal). In this case, Paul Lukacs (now with Colorado Division of Wildlife) has developed theory and models in MARK to deal with this, in the context of genotyping errors. If this isn't the case, the problem is much harder.

Cheers, Paul

[cschwarz@stat.sfu.ca]

If you are particularly interested in effects upon abundance (e.g. is the populacion declining), try looking at:

Schwarz C. J. and Stobo W. T. (1999). The effect of and estimating misread rates in capture-recapture studies. Canadian Journal of Fisheries and Aquatic Sciences 56, 551-559.

McKelvey K. S. and Schwartz M. K. (2004). Genetic Errors Associated with Population Estimation Using Non-invasive Molecular Tagging: Problems and New Solutions. Journal of Wildlife Management 68, 439-448.

Lee A. (2002). Effect of List Errors on the Estimation of Population Size. Biometrics 58, 185-191.


As mentioned earlier, Paul Luckas has done some nice work on genetic methods.

[Paul Lukacs]

Errors in genetic studies differ from those in tagging studies a bit because in tagging studies you should know what marks you've placed in the population. In genetic studies, all animals have a genotype and the researcher is trying to determine what that genotype is. Typically, with genetic errors they appear as "new" individuals in the population whereas with a tag reading error you are more likely to swap individuals because you know immediately when you read a tag that isn't actually in the population. Therefore, the misidentification models in MARK may not apply to your situation.

Paul Lukacs

[b06jll]

In response to Paul's note, the probability of mis-IDing an individual and recording it as another (real) individual is very high. Of all the errors recorded during my study, 94% correctly corresponded to another individual in the banding dataset - not surprising given that we have more than 5,000 Razorbills banded in a small area.
I appreciate everyones input.
Jennifer