Use recapture overlap while recognizing weak identification

Define the ecological target

Mark–recapture methods use the overlap between samples to learn about individuals not observed. The basic idea compares the marked fraction in a later sample with the marked fraction of the population. This requires a defined closed population and a sampling process that makes the marked individuals appropriately representative.

Specify the observation model

The supplied first sample marks M=20 individuals. A second sample captures C, of which R are marked recaptures. The number of distinct observed individuals is M+C−R; subtract overlap once. The Chapman corrected estimate is ((M+1)(C+1)/(R+1))−1. It is an estimate, potentially fractional, not a literal partial animal or a census.

Work through the sampling logic

The model also supplies a conventional variance estimate: (M+1)(C+1)(M−R)(C−R)/((R+1)²(R+2)). Its square root has individual units, unlike variance. These expressions rely on the sampling assumptions and do not guarantee a reliable interval at very low overlap. The interface intentionally does not report a normal confidence interval that could be misleading.

Challenge the inference

When R=0, the corrected formula is finite, but the data have little leverage to distinguish a large population from poor mixing or low capture probability. A finite formula is not proof of a well-identified finite upper bound. High capture heterogeneity, mark loss, behavioral response, movement and mortality can also undermine the model.

Design a defensible survey

This is a virtual count exercise, not permission to trap, tag or release animals. A real study needs qualified ethical and legal review, appropriate methods and a model matched to its design. Preserve capture histories and identification uncertainty rather than reduce everything prematurely to a single point estimate.

Worked original packet

With M=20, C=40, R=8, observed distinct individuals are 52. Chapman estimate is 21×41/9−1≈94.67. The model variance is 21×41×12×32/(81×10)≈408.18, giving standard error about 20.20 under the stated assumptions.

Investigation

Compare overlaps 0, 4 and 8 at second count 40. Track distinct observed individuals, estimate and model standard error, then explain a non-abundance mechanism that could reduce overlap.

Changed-case transfer

For M=20, C=30, R=10, find distinct observed count and the corrected estimate. Why is observed distinct count not the same as estimated total?

Feedback

Distinct observed 40. Estimate 21×31/11−1≈58.18. The estimate allows for individuals never captured, conditional on closure and comparable capture; it is not guaranteed to equal the actual population.

Sources and evidence boundary

Original instructional prose, values and exercises. Context checked 30 October 2026: primary agency source. All observations are synthetic. No organism collection, trapping, tagging, release or unsafe-site access is requested; a real study requires qualified design, permission and supervision. Public feedback is practice, not independently verified mastery.