We conducted a study in order to determine the factors that influenced turtle nest site selection. We collected various environmental variables from turtle nests, as well as three controls (i.e. locations not selected by turtles for nests). Positive controls (PC) were placed within 10m of a turtle nest, within the same habitat patch (e.g. a nest bank). Negative one controls (NC1) were placed on a similar habitat patch, but with no turtle nests present. Negative two controls were placed within 300m of a turtle nest, at a random angle (compass randomly spun) and random distance (random number table) from said nest. The controls were designed with these three basic questions in mind: 1) Why did turtles chose a nest patch over any other available habitat patch (i.e. compare data from NC2's to data from NC1's and Nests); 2) Why did turtles chose a nest patch over another similar nest patch (i.e. compare data from NC1's to data from Nests); and 3) Within a patch, why did a turtle chose this specific location to nest (i.e. compare data from PC's to Nests).

As one would expect, the data collected from the nests, PC's, and NC1's were often similar (e.g. similar bank slope); While data from NC2's were very different.

We fitted the data to a binomial, generalized linear model in which nest selection was the predictor variable (i.e. nests are 1's and all controls were 0's). (Models with all possible variable combinations examined and models ranked using AICc)

We have run into a problem where the similarities between the nests and PC's and NC1's seem 'outweigh' or overwhelm the variation from the NC2's. For example, we expected dominant aspect to be an important contributor to turtle nest selection (this is based upon our literature review, and exploratory anova analysis that the mean aspect between nests and NC2's were very different), however, it was not included in our top model.

Should we consider a structured modeling approach? Or, should we instead just run three separate models and report the results at the three different scales (i.e. nest & PC, nest & NC1, nest & NC2)?

Thank you for any help provided! I can provide additional information or data (raw data or glm results) if necessary.