Known Fate
Known Fate
The data coding for the known fate model requires a 1 in the L part of the encounter history for every occasion that the animal is alive at the start of the interval and its fate is known through the interval. A 10 means the animal lived through the interval, and a 11 means the animal died during the interval. There is no code of 01 allowed in the known fate model — this means that the animal was not alive at the start of the interval, so could not have died. To censor an animal for an occasion when you don’t know the animal’s status actually requires censoring an animal for both the preceeding and following intervals. Censoring is specified with the 00 code. Thus, the encounter history 00101000001011 means that the animal lived through intervals 2 and 3 (because he animal is known alive at tme 4 meaning it survived interval 3), was censored for interval 4 and 5 because you did not know the animal’s status at time 5. Then, the animal was found alive during the 5th interval (i.e., known alive at time 6, the start of the 6th interval), and then lived through interval 6, and died in interval 7.
Note that the Known Fate model is different than the Nest Survival model.
A individual random effect can be added to the Known Fate model by using the Change Data Type menu choice. However, the random effects sigma parameter will be non-identifiable if you have constructed a saturated model. For Known Fate data, the saturated model would be unique parameter for each group*time combination when there are no individual covariates. In such cases, the -2log Likelihood of the model you have constructed is the same as the saturated model, so that the deviance is zero, and the random effects sigma parameter will be non-identifiable.
One of the main purposes of the random effects known fate data type is for simulation, where data with overdispersion (i.e., parameter heterogeneity) can be generated and used in the usual known fate model to evaluate the effect of parameter heterogeneity.