Using observation-level random effects to model overdispersion in count data in ecology and evolution

By X. Harrison. PeerJ, 2014

1,065 citations recorded by the source corpus.

Overdispersion is common in models of count data in ecology and evolutionary biology, and can occur due to missing covariates, non-independent (aggregated) data, or an excess frequency of zeroes (zero-inflation). Accounting for overdispersion in such models is vital, as failing to do so can lead to biased parameter estimates, and false conclusions regarding hypotheses of interest. Observation-level random effects (OLRE), where each data point receives a unique level of a random effect that models the extra-Poisson variation present in the data, are commonly employed to cope with overdispersion in count data. However studies investigating the efficacy of observation-level random effects as a means to deal with overdispersion are scarce. Here I use simulations to show that in cases where overdispersion is caused by random extra-Poisson noise, or aggregation in the count data, observation-level random effects yield more accurate parameter estimates compared to when overdispersion is simply ignored. Conversely, OLRE fail to reduce bias in zero-inflated data, and in some cases increase bias at high levels of overdispersion. There was a positive relationship between the magnitude of overdispersion and the degree of bias in parameter estimates. Critically, the simulations reveal that failing to account for overdispersion in mixed models can erroneously inflate measures of explained variance (r2), which may lead to researchers overestimating the predictive power of variables of interest. This work suggests use of observation-level random effects provides a simple and robust means to account for overdispersion in count data, but also that their ability to minimise bias is not uniform across all types of overdispersion and must be applied judiciously.

Subfields: Ecological Modeling, Ecology, Evolution, Behavior and Systematics, Nature and Landscape Conservation

View Paper: https://www.semanticscholar.org/paper/4f2ae28fb82e870b65f8d02480edf58b5ae7eacf

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Ecological Modeling

In Ecology, Evolution, Behavior and Systematics

In Nature and Landscape Conservation

Citations in the corpus

The coefficient of determination R2 and intra-class correlation coefficient from generalized linear mixed-effects models revisited and expanded (2016, positive): “Here, overdispersion 84 represents the excess variation relative to what is expected from a certain distribution and can be 85 estimated by fitting an observation-level random effect (OLRE; see, [12, 13]).”
glmmTMB Balances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling (2017, background): “Models that ignore zero-inﬂation, or attempt to handle it in the same way as simple overdispersion, can yield biased parameter estimates (Harrison, 2014).”
The consequences of checking for zero‐inflation and overdispersion in the analysis of count data (2019, background): “If model residuals are overdispersed or have an excess of zeros, assumptions underlying a Poisson GLM will not hold and ignoring this will lead to serious errors (e.g. biased parameter estimates and invalid standard errors; Harrison, 2014).”
Functional traits and community composition: A comparison among community‐weighted means, weighted correlations, and multilevel models (2018, background): “and the Poisson distribution to allow for overdispersion (Harrison 2014).”
Modeling zero-inflated count data with glmmTMB (2017, background): “Models that ignore zero-inflation, or attempt to handle it in the same way as simple overdispersion, yield biased parameter estimates (Harrison, 2014).”
A comparison of observation-level random effect and Beta-Binomial models for modelling overdispersion in Binomial data in ecology & evolution (2015, methodology): “Harrison (2014) also provides code to estimate the dispersion parameter and 95% confidence intervals using parametric bootstrapping.”
A brief introduction to mixed effects modelling and multi-model inference in ecology (2018, methodology): “The R package ‘sjPlot’ (Lüdecke 2017) (Harrison 2014; 2015).”
Ants as indicators of environmental change and ecosystem processes (2017, methodology): “In ddition, we included a random factor for each observation event to emove overdispersion (Harrison, 2014).”
Biodiversity along temperate forest succession (2018, methodology): “To account for the overdispersion frequently observed in models of count data, we included an observation- level random effect (i.e., the plot) in all models (Harrison, 2014).”
The indicator side of tree microhabitats: A multi-taxon approach based on bats, birds and saproxylic beetles (2018, methodology): “We added an observation- level random effect to account for potential overdispersion (Harrison, 2014).”