Dear Alessandro, thank you for your pointer. This would work well if my observations were perfect.

I struggle with the uncertainty of the observations. The food source is often not reported. On the mollusca I am more certain but also not 100%. Would there be a way to incorporate these uncertainties. I think I have a different way of estimating these. 

Regards, Leon

Hi Leon Hoffman.

Since you do not have direct evidence on the relationship between carnivores and its food source, I suggest that maybe you could make an approach to the "Community interaction web" using Path Analysis. In the link you will find a reference: 

http://www.jstor.org/stable/1939391?seq=1#page_scan_tab_contents

Dear Leon,

I would suggest you to use Structural Equation modeling. This method allows you to build and to test your relationship network. Thus, it could allow you to test whether your food web fit the data you have.

Jim Grace wrote a lot about SE models applied to ecology and analysis of complex natural systems. You can contact him for sure. 

https://www.researchgate.net/profile/James_Grace

Hope it helps.

Cheers,

Guillaume

Leon:

I recomend the next 3 papers about a proved method:  EcoPath with Ecosim

 Christensen, V., Walters, C., 2004. Ecopath with Ecosim: methods, capabilities and limitations. Ecological Modelling 172, 109–139.

Christensen, V., Pauly, D., 1992. ECOPATH II A. Software for balancing steady state ecosystem models and calculating network characteristics. Ecological

Modelling 61, 169–185.

Christensen, V., Walters, C.J., Pauly, D., 2005. ECOPATH with ECOSIM: a User’s Guide. Fisheries Centre, University of British Columbia, Vancouver (November 2005  ed., 154 pp).

They are cited by Neira et al in : Mechanisms affecting recovery in an upwelling food web: The case of the southern Humboldt, published by Neira et al in Progress in Oceanography 83 (2009) 404–416.  Thanks to Francisco Fernandez from Sernapesca to help answer.

Leon:  I just find another seminal paper abnout your question, I hope will be of utility.

ICES Journal of Marine Science, 57: 697–706. 2000 doi:10.1006/jmsc.2000.0726, available online at http://www.idealibrary.com on Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries

Daniel Pauly, Villy Christensen, and Carl Walters

Pauly, D., Christensen, V., and Walters, C. 2000. Ecopath, Ecosim, and Ecospace as tools for evaluating ecosystem impact of fisheries. – ICES Journal of Marine Science,

57: 697–706.

Regrettably, you will not find the statistical method to prove that one species feed on another. However, You may try to demonstrate the co-occurrence of both species, assuming that such a co-occurrence might imply some relationship between these two species in a trophic chain. To evaluate the co-occurrence ypu may try to employ the log-likelihood test (see the enclosed paper for a discussion). You may also calculate the conditional probabilities, considering that the incidences are different for species A and B. You may use Q Dixon test for two groups to show the parallelism, etc. However, even if you reject the null hypothesis stating that two species co-occur by pure chance in your observations, it is still far from proving that one species feeds on another. For this purpose yu should rather look for molecular evidence.