| Popis: |
Datastream permutations are commonly used to test null hypotheses in animal social network analysis. Permutation methods are inherently stochastic and, like all stochastic processes, can be unreliable if appropriate diagnostic procedures aren't employed. Though datastream permutations are widely used in behavioural ecology, sufficient diagnostic checks have not yet been adopted to guarantee their reliability. In this paper we highlight that without proper checks, datastream permutations can be severely unreliable, but that using diagnostic tools developed for Markov chain Monte Carlo methods can improve the reliability of inferences. |
