Environmental Engineering Reference
In-Depth Information
A more flexible approach to analysing known fate data is to use 'time-to-event'
survival analyses
such as the Cox proportional hazards, Kaplan-Meier or para-
metric survival models (Smith 2005; StatSoft n.d.). These approaches estimate
mortality rates from the precise times between marking and death for a sample of
individuals. One benefit of this approach over the life table is that it is easy to
include in the analysis individuals that are followed for a period but not observed
to die before monitoring ceases, so long as a reasonable number of deaths are
actually observed. More importantly, these methods are forms of regression
analysis that allow the influence of possible explanatory factors on survival rate to
be formally tested. In common with mark-recapture methods, survival rate
estimates based on known fates are only
representative
of the wider population
to the extent that the individuals observed are representative. The usual care in
selecting study subjects is therefore required.
2.4.2.3 Which method is best?
The range of methods available for the estimation of survival is as wide as that for
abundance estimation; however, the choice of method will usually be much more
restricted by the type of data that can be obtained from your organism
(Figure 2.12). Where there is a genuine choice, individual-based methods should
be preferred, since they generally provide the most robust way to obtain unbiased
estimates of survival, and allow testing of hypotheses about the factors that may
influence it (such as harvest pressure).
Type of
data
Known
individuals
Counts
Type of
count
Certainty of
follow-up
Full age
structure
sampled
Certain or
Near certain
Uncertain
Numbers of
young/adults
Type of
follow-up
Known fate
Static
life-table
Population
ratios
Recovered
dead
Resighted or
recaptured live
Frequency
of follow-up
CJS
mark-recapture
Dead
recoveries
Continuous or
near-continuous
Periodic
Cohort
life-table
Survival
regression
Fig. 2.12
A decision tree for identifying the most appropriate survival estimation
method for a given type of data.
