Capture recapture: developments and applications and 2010 AGM
Friday 26th November 2010
The meeting is joint with the Royal Statistical Society Young Statisticians Section, and will be preceded by a two-hour tutorial on capture-recapture.
BIR members: £10 for lunch, otherwise free.
BIR student members: £5 for lunch, otherwise free.
Non-members: tutorial, lunch and afternoon meeting £55; Lunch and afternoon meeting £45; afternoon meeting only £35.
Email email@example.com with the subject header REGISTRATION before Nov 5th, indicating whether you are a member and which parts of the day you wish to register for. Sue will reply to confirm the fee and the address for posting cheques, which should be made out to the Biometric Society.
Non-members are encouraged to joint the Society (membership information ) to take immediate advantage of the members' registration fee.
Instructions on how to reach Rothamsted
|Document downloads for IBS members.|
Join us now.
|10.00 - 10.30||Tutorial Registration and coffee|
|10.30 - 12.30||Introduction to methods of Capture-recapture and their application|
This is an optional tutorial by Rachel McCrea and Byron Morgan, University of Kent.
We start by describing the history of capture-recapture methods, as they have developed over the past 100 years. We then describe the wide range of extremely useful procedures currently available, for estimating population sizes, as well as the survival and movement of wild animals. In addition, it will be explained how the methods may be applied more widely, for example to medical and social data. The tutorial will also describe the different specialist computer packages that now exist for the analysis of capture-recapture data. Throughout there will be a range of real-life illustrations. No previous knowledge of the area is required in order to attend this tutorial.
|12.30 - 13.20||Lunch |
This is an optional sandwich lunch.
|13.20 - 13.45||Annual General Meeting |
|13.45 - 14.30||IBS-BIR Presidential address: Of mice and men: recent developments in capture-recapture|
Byron Morgan, University of Kent
The first methods for the analysis of capture-recapture data were based on simple probability models designed for small data sets. The situation today is far more complex and interesting. This is due in part to the development of new technology, which has resulted in sophisticated ways to mark and track wild animals. In addition, long-term data sets are now becoming available, as well as very detailed studies of particular animal populations, including Soay sheep, red deer and meercats. On the modelling front, new research has resulted in a better understanding of how different types of models are connected. In this talk several of these unifying approaches will be described, including state-space models, multi-state systems, stop-over models and hidden Markov models. Links with models for human survival and demography will be made, and the talk will be illustrated by a range of real data.
|14.30 - 15.00||Boosting qualifies capture- recapture methods for estimating the comprehensiveness of literature|
Gerta Rucker, Freiburg University
We investigate the feasibility of capture- recapture techniques in combination with model selection for estimating the number of missing references in literature searches, using two systematic reviews in gastroenterology and hematology. We compared manually-selected Poisson regression models differing with respect to included interactions, and performed model-selection via componentwise boosting, which provides automatic variable selection. The technique used is a regularized, stepwise procedure, allowing distinction between mandatory and optional variables. For the first example, the most plausible manually-selected model suggested a number of 82 missing articles (CI 95% [52;128]), while the boosting technique provided 127 (CI 95% [86;186]) missing articles. For the second example, the values were respectively 140 (CI 95% [116;168]) and 188 (CI 95% [159;223]. Boosting is found to be robust against overfitting, and automatically creates a proper model for inference.
|15.30 - 15.50||Capture-Recapture Inference under Structured Heterogeneity|
Dankmar Bohning, University of Reading
It is well-known that inference in the capture-recapture setting leads to biased inference when heterogeneity occurs and estimators used were developed under homogeneity. Up to today no valid estimators of the population size exist when heterogeneity is present. The talk suggests using adjustments of the popular Chao- and Turing-estimators, which are based on structured heterogeneity, a concept introduced to characterize a certain distributional pattern in the latent variable which can be diagnosed via a simple graphical tool. This simple graphical tool, called the ratio plot, can be used to diagnose the presence of structural heterogeneity. The structural heterogeneity is characterized by a parameter which can be estimated via a regression technique or alternatively by means of an EM algorithm. Simulation studies show that the adjustments based upon incorporating structural heterogeneity leads to asymptotically unbiased estimators. An analysis of a variety of examples shows that almost always heterogeneity occurs in real life applications of capture-recapture studies. However, as the talk will also demonstrate, the heterogeneity is often structured so that it can be described by a simple distributional model. Estimating the parameter in the model characterizing the structured heterogeneity leads to an adjustment which results in asymptotically unbiased estimators - as the simulation study shows. Applications to real life capture-recapture data illustrate the improvements.
|15.50 - 16.20||Bayesian Capture-Recapture Analyses Using a State-Space Framework|
Ruth King, University of St Andrews
Recent interest in the analysis of ecological capture-recapture data has included the use of state-space modelling within a Bayesian framework. We review the use of a state-space modelling framework for capture-recapture data and demonstrate the flexibility by considering a number of extensions to the basic Cormack-Jolly-Seber model. For example, these include integrated recapture-recovery data, multi-state data, partially-observed data, individual continuous covariates, individual random effects, stopover data and closed populations. We show that these potentially complex models can be easily fitted within a Bayesian framework using the freely available computer program WinBUGS, providing a mechanism for ecologists to fit these models. Finally we provide a short discussion of the advantages and disadvantages of the Bayesian state-space approach.
|16.20 - 16.40||Invited discussion by Bill Browne (University of Bristol) |
|16.40 - 15.00||General Discussion|
Existing members can login below to view all site content. Lost password?
Other visitors might be interested to learn more about the benefits of membership.