Where does overdispersion collide?

In spatial ecology, a pattern is overdispersed when individuals are spaced more evenly than they would be at random: a regular, repulsive arrangement, as when territorial plants or animals hold one another at a distance. Clark and Evans' 1954 Distance to nearest neighbor as a measure of spatial relationships in populations gave the nearest-neighbour index that separates overdispersed (regular) patterns from random and clumped ones. Confusingly, in parts of the ecological literature the word means less clumping and so less variability in local counts, the near-opposite of what the same word denotes in statistics.[1]

Unit

How regular versus clumped the spacing of individuals is

Recovery means

Not applicable

Measured by

Quantified by the Clark-Evans nearest-neighbour index, where a ratio above one indicates overdispersion toward regularity

In statistical modelling, overdispersion is the condition in which observed data show more variability than the fitted model allows (most often counts whose variance exceeds the Poisson mean, or proportions more variable than the binomial predicts), usually because of unmodelled heterogeneity or clustering. Cox's 1983 Some remarks on overdispersion set out its causes and consequences and how to adjust inference for it. Here overdispersion means more variance, not more even spacing, so a spatially overdispersed (regular) population would in count terms actually be under-dispersed, the same word pointing opposite ways.[2]

Unit

How far the data's variance exceeds the model's assumed variance

Recovery means

Not applicable

Measured by

Detected by a variance-to-mean ratio or a deviance that exceeds its degrees of freedom, and corrected with a dispersion parameter