Extended Data Table 1 Model selection

From: Global status and conservation potential of reef sharks

  1. Model selection for the candidate negative binomial models used for the estimation of the relative abundance of reef sharks (MaxN) within the Global FinPrint dataset. The negative binomial model used was parameterized as a Poisson random variable for which the rate parameter is gamma-distributed. ZI, zero-inflated; LOO, leave-one-out cross-validation; SE, standard error of the LOO; ΔLOO, the net difference in LOO scores relative to the LOO for the lowest-scoring model; weight, the probability of each model being equivalent to the lowest-ranking LOO model (ZI full), based on the standard error of the difference in LOO scores (see Methods).
Back to article page