How to register a captured, non marked individuals

[Jonathan]

Hello,

I hope I'm not missing something overtly obvious here but:

I'm trying to run a comparison with a mark - recapture analysis published by Rivot and Prévost (2002) - full reference below.

Their analyses includes capture-mark-recapture models for both salmon smolts and spawners. To initially focus on just the smolts, the data they use includes:
• "ci" the number of smolts caught;
• "mi" the number of smolts marked and released; and
• "rmi" the number of marked recaptured

So here's my question, how do I introduce to MARK data identifying fish which were caught in the first instance, not marked, but released.

As an example,
We have two encounter occasions:
• 10 represents marked individual which are caught, marked but not re-caught,
• 11 individuals which are caught, marked and re-caught,
• 00 individuals which are caught, not marked and hence not re-caught

Here's the catch - a "00" entry caries no information and the analysis (expectedly) encountes errors. How do I introduce the information "caught, not marked, released" as this has an effect on the capture efficiency estimates...

Having run without the “00” entities I can re-create the trends estimated by Rivot and Prévost, but get markedly different value estimates.

Am I missing a model type in MARK or is it not possible to take non-marked captures into consideration for estimating capture efficiency?

Any views, ideas or advice?
And as initially stated – I hope I’m not missing something too obvious.

Cheers,

Jonathan.

Ref:
Rivot, E., and Prévost, E. 2002. Hierarchical Bayesian analysis of capture–mark–recapture data. Can. J. Fish. Aquat. Sci. 59: 1768–1784.

[cschwarz@stat.sfu.ca]

If you are interested in Cormack-Jolly-Seber type models (i.e. only interested in survival and capture) the unmarked, captured individuals are non-informative and can be discarded. The CJS models estimate the survival and capture rates of marked individuals only -- the key assumption is that these are applicable to the unmarked segment of the population.

You state:
"00 individuals which are caught, not marked and hence not re-caught :"
How do you know that these individuals are not recaught if they are not marked? If these are individuals that are captured and lost on capture (ie. removed as harvest), then they are not released back to the pool of marked fish and again contribute no information on the subsequent survival and capture rates.

If you are interested in estimating abundance (Jolly-Seber models), then captures of fish that are unmarked are used to estimate abundance (typically $\widehat{N}_i = \frac{n_i}{\widehat{p}_i}) and so are used then. But again, these fish contribute no information about the marking efficiency.


The Rivot and Prevost paper use a Bayesian method that "shares" information across years. Hence information about capture-efficiency in a particular year is also influenced by information about capture-efficiency in other years. This is likely why your estimates differ substantially.

[Jonathan]

Thanks for the quick, and informative, reply,

This is interesting and confirms my suspicions that: a) I might have missed the obvious (!) but most importantly; b) individuals caught in an initial capture instance and released un-marked, cannot be included or contribute toward estimations.

Again – thanks for bearing with me, it’s been a while and a 700+ page manual takes a little digesting, even as nicely written as it is.

Cheers,

J.