(Beintema and Muskens 1987; Ens 1991), but it is often the case that relatively
few nests are found during egg-laying, so the evidence for a difference is weak and
the temptation is to combine the two periods.
Statistical tests of differences in Mayfield daily nest failure (or survival rates)
among breeding stages, study areas or years can readily be conducted by treat-
ing each exposure day as if it was a binomial trial during which the nest can
fail or not. Standard errors of the daily probability of success or failure can be
calculated by the method of Johnson (1979). Daily failure probabilities can
be modeled using logistic regression with the generalized linear modeling
facilities available in many statistical packages (Etheridge et al. 1997; Aebischer
1999) and likelihood-ratio tests used to identify a minimal adequate model
(Crawley 1993). This has the advantage that failure probability can be modeled as
a function of several variables measured around nest sites and these can include
both continuous variables (such as nest height) and categorical variables (such
as nest tree species). It is often desirable to know the overall probability of nest
survival over a series of successive breeding stages, such as egg-laying, incuba-
tion, and the nestling period, for which separate Mayfield estimates have been
calculated. This survival rate is simply the product of the rates for the successive
stages. The standard error and confidence limits of the product can be calculated
using methods described by Hensler (1985). This product can be multiplied by
the mean number of young fledged per successful nest to give an estimate of the
mean number of young fledged per breeding attempt. Rotella et al. (2000)
described a method that could be used to estimate both the daily nest failure rate
and any adverse effect on it of nest checking. The method uses the assumption
that any effect of checking would result in failure soon after the check, so longer
intervals between checks yield lower apparent daily failure rates than short
intervals.
3.4 Determination of the proximate causes of breeding failure
3.4.1Signs left at the nest
The evidence left at the nest may be helpful in determining the proximate
cause of failure of a breeding attempt, but assessments of this type are likely to
be open to error. Consider a hypothetical case in which there are two species of
egg predator that each depredate half of the egg-stage breeding failures of the
study species. Both predators leave characteristic signs that allow each predator
Determination of the proximate causes of breeding failure| 67