Objectives

The purpose of this exercise is to demonstrate how mark-recapture data are collected and recorded for a simple study such as might be done for small mammals. Then to see how reliable the inferences made from the data by MARK or CAPTURE prove to be when compared with the known parameters of the population.

MARK uses maximum likelihood estimators to obtain the best estimates of the model parameters, then compares the models using AIC. You should be familiar with these concepts already: check the units "Frogs in ponds - maximum likelihood estimators" and "More frogs in ponds - AIC and likelihood" if necessary.

Materials needed

For this experiment you’ll need approximately 50 identical cards, such as business cards or index cards.

You should also print out the data sheet which is the last page of the lab guide.

Working through the experiment

Download the lab guide "Mark-recap_cards.pdf". You will probably want to print out the lab guide and have it next to your computer while you work through the instructions.

Work through the lab guide before checking the results below.

Results

As well as looking at the AICc values and the estimated population, carefully check the results for probability values which are near zero or near 1, and standard errors which are near zero. These usually indicate that MARK has not been able to estimate the parameter properly, but are sometimes a reasonable inference from the data, for example, if no animals at all were captured during one trapping session it would be reasonable to infer that the capture probability for that session was 0.

If you "captured" the suggested number of cards (20 the first time, then 16, 22, 8 and 20) the capture probability varied widely between trapping occasions. Since there is no difference between the cards (no heterogeneity) and no difference between the first capture and recaptures, the correct model is Mt.

In some cases you will find that random sampling error obscures the time effect, and the M0, Mb or Mh models may sometimes appear better.

Most runs of this experiment produce results which are surprisingly far from the true value - at least, it always surprises me! This is a reminder that we are producing imprecise estimates and we should take the confidence intervals seriously.

Generally the true value will lie within the confidence interval, but sometimes (theoretically about 5% of runs) it will be outside the confidence interval.

Even with no issues with data collection or with meeting the assumptions of the analysis method, our results are not as precise as we might have imagined!

If you want to explore simulated capture histories without having to work with cards, you will find an R function and script here.