A brief guide to model selection, multimodel inference and model averaging in behavioural ecology using Akaike’s information criterion

By Matthew Symonds, A. Moussalli. Behavioral Ecology and Sociobiology, 2010

2,414 citations recorded by the source corpus.

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

View Paper: https://www.semanticscholar.org/paper/28a33248a3825855a7cd8d80c2c361d0e8d56753

How researchers see this paper

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

In Ecology, Evolution, Behavior and Systematics

In Nature and Landscape Conservation

Citations in the corpus

Issues in information theory-based statistical inference—a commentary from a frequentist’s perspective (2010, methodology): “…first the full model is tested using classical NHST, and once this revealed significance, the most parsimonious model (or a set of models) is selected using an IT-based approach (see also Symonds and Moussalli 2010; for more reasons to inspect the full model, see Forstmeier and Schielzeth 2010).”
Multimodel inference in ecology and evolution: challenges and solutions (2011, methodology): “…when focussing on linear regression-type analysis, AICC remains the most widely used criterion; it is also the most easily applied because it is implemented in model averaging packages in R [such as MuMIn (Bartoń, 2009)] and most other major statistical packages (Symonds & Moussalli, 2010).”
Ecologists overestimate the importance of predictor variables in model averaging: a plea for cautious interpretations (2014, methodology): “Contrary to AIC, AICc includes a correction for small sample size and is now recommended over AIC (Burnham & Anderson 2002, 2004, Symonds & Moussalli 2011, but see Richards 2005, Turek & Fletcher 2012).”
Black, Grey and Watch Lists of alien species in the Czech Republic based on environmental impacts and management strategy (2016, background): “The evidence ratio between these two models suggested the highest-ranked model was 1.9 times more likely to be the most parsimonious model than the secondranked model, but a Δ AICc 2 indicated that the two models were nearly equivalent (Symonds and Moussalli 2011).”
A farewell to the sum of Akaike weights: The benefits of alternative metrics for variable importance estimations in model selection (2017, methodology): “As a consequence, SW is also largely defined as the probability that the variable of in - terest is included in this best model, independent of other variables in the set (Symonds & Moussalli, 2011).”
A brief introduction to mixed effects modelling and multi-model inference in ecology (2018, background): “Further reading: Grueber et al (2011) and Symonds & Moussalli (2011) give a broad overview of multi-model inference in ecology, and provide a worked model selection exercise.”
Evaluating a coupled discrete wavelet transform and support vector regression for daily and monthly streamflow forecasting (2014, methodology): “In the AIC scheme, the (Akaike) weight xm for a given AIC model m is estimated as follows (Buckland et al., 1997; Burnham and Anderson, 2002; Symonds and Moussalli, 2011):”
Model averaging in ecology: a review of Bayesian, information-theoretic, and tactical approaches for predictive inference (2018, background): “…Omland 2004, Hobbs and Hilborn 2006, Burnham et al. 2011, Freckleton 2011, Grueber et al. 2011, Nakagawa and Freckleton 2011, Richards et al. 2011, Symonds and Moussalli 2011), probably under the influence of the AIC-weighted (Akaike information criterion) averaging popularized by Burnham and…”
Using information theory as a substitute for stepwise regression in ecology and behavior (2010, neutral): “The special issue to which our work belongs (see Garamszegi 2010) contains papers that generally describe IT methods in either technical (Burnham et al. 2010) or application-oriented ways (Symonds and Moussalli 2010).”
Model averaging, missing data and multiple imputation: a case study for behavioural ecology (2010, methodology): “As a result, model averaging offers more reliable and robust point and uncertainty estimation of parameters (for more detailed descriptions model averaging, see Burnham and Anderson 2002; Richards et al. 2010; Symonds and Moussalli 2010).”
Testing Fit Patterns with Polynomial Regression Models (2015, methodology): “If ∆AICc is larger than 10, then the model is considered to be implausible compared to the best model (Symonds & Moussalli, 2011; Burnham et al., 2011).”
A meta-analysis of correlated behaviors with implications for behavioral syndromes: relationships between particular behavioral traits (2013, methodology): “, effect sizes and CIs) based on the relative fit of the models including such terms (Symonds and Moussalli 2011).”
Quantifying variable importance in a multimodel inference framework (2016, methodology): “This benchmark is inappropriate because it is not the quantity that SW is defined to estimate; instead, SW reflects the probability that a given variable is included in the best model based on information theory (Symonds & Moussalli 2011).”