To determine if the proportion of prawns selecting a given habitat varied with any of the factors investigated, log-linear analyses were used.
Each combination of habitats was analysed separately.
Initially, a single analysis was done incorporating both the three different size classes and two predator types, with the individual factor habitat, and all interactions between habitat and time (day/night and size/predator (small, medium and large with no predators, medium + mullett, medium +crab).
Time and size/predator and their interaction have zero deviance by definition as all animals had to select one of the two habitats.
Significance was determined using an analysis of deviance testing against the X2 distribution (akin to ANOVA with continuous data, Chambers and Hastie, 1993).
If size/predator interacted with habitat (indicating that it had an influence on what habitat was selector), then separate analyses were conducted for the three size classes without predators, and for medium prawns under different predation regimes, to determine which factor was important (note that medium prawns without predators appeared in both of these later analyses).