Dealing with collinearity in behavioural and ecological data: model averaging and the problems of measurement error

By R. Freckleton. Behavioral Ecology and Sociobiology, 2010

339 citations recorded by the source corpus.

Subfields: Ecology, Evolution, Behavior and Systematics, Small Animals, Statistics and Probability

View Paper: https://www.semanticscholar.org/paper/60211163b32fee681e5aa79f91480d66cb65746c

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 Small Animals

In Statistics and Probability

Citations in the corpus

Model averaging and muddled multimodel inferences. (2015, methodology): “Similar concerns apply to simulations done by Freckleton (2011) for evaluating the bias of model-averaged coefficients with increasing multicollinearity among predictors.”
Information-theoretic approaches to statistical analysis in behavioural ecology: an introduction (2010, background): “Nakagawa and Freckleton (2010) investigate the effect of missing observations on the IT output in an example analysis.”
Ecologists overestimate the importance of predictor variables in model averaging: a plea for cautious interpretations (2014, background): “Every variable was normally distributed, and there was no evidence for collinearity between predictor variables (Freckleton 2011), as indicated by the maximum value of variance inflation factor of 1 1 (O’Brien 2007).”
Statistical Issues and Assumptions of Phylogenetic Generalized Least Squares (2014, background): “4 Note that residuals of a PGLS are actually multivariate normal (Freckleton et al. 2011), which has implications for practical checks of their distribution; see the Online Practical Material (http://www.”
Model averaging, missing data and multiple imputation: a case study for behavioural ecology (2010, background): “These can include the possible correlations between the variables, which will be determined by the pattern of missingness, with consequences for parameter estimates and variances (Freckleton 2010).”
A brief introduction to mixed effects modelling and multi-model inference in ecology (2018, methodology): “Nakagawa & Freckleton (2011) recommend multiple imputation (MI) as a mechanism for handling missing data, and highlight the ability of this technique for more accurate estimates, particularly for IT-AIC approaches.”
Issues in information theory-based statistical inference—a commentary from a frequentist’s perspective (2010, background): “When collinearity does exist, results tend to be unstable (small differences in the analysed data may lead to large changes in the parameters estimated), and estimates of parameters have large standard errors (implying that the estimated effect of a predictor variable is associated with large uncert”
Here be dragons: a tool for quantifying novelty due to covariate range and correlation change when projecting species distribution models (2014, background): “For example, correlation coefficients as high as Pearson’s r = 0.8 are common in the literature (Elith et al., 2006, 2010), and indeed, (multi)collinearity is considered as a serious challenge to most of the correlative modelling methods (Graham, 2003; Freckleton, 2011; Dormann et al., 2013).”
Nest visit synchrony is high and correlates with reproductive success in the wild Zebra finch Taeniopygia guttata (2012, background): “4 was unlikely to have affected our results (Freckleton 2010).”
Model selection and model averaging in behavioural ecology: the utility of the IT-AIC framework (2010, methodology): “Further problems of stepwise approaches are that they can only be used to choose among a set of completely nested models and that they seldom deal well with collinear variables (Freckleton 2010).”
AIC model selection and multimodel inference in behavioral ecology: some background, observations, and comparisons (2011, methodology): “We omitted additive models including both of these variables together in the same model because we think they are redundant (see e.g., Freckleton 2010 ) Lastly, we considered interactions between pairs of the variables and added to our model set interactive models of body size with male dominance s”
Quantifying variable importance in a multimodel inference framework (2016, background): “…ofR2 among correlated variables when calculating Iweighted would be to consider them in tandem when constructing candidate model sets (i.e. correlated variable pairs are always entered into a model together; Freckleton 2011 first suggested this strategy to reduce bias in parameter estimates).”
Using information theory as a substitute for stepwise regression in ecology and behavior (2010, neutral): “2010), and dealing with collinearity and missing observations during this process (Freckleton 2010; Nakagawa and Freckleton 2010).”
Model averaging in ecology: a review of Bayesian, information-theoretic, and tactical approaches for predictive inference (2018, background): “Several previous reviews on model averaging in ecology and 53 evolution, focussed exclusively on ‘information-theoretical model averaging’ (Johnson 54 and Omland, 2004; Hobbs and Hilborn, 2006; Burnham et al., 2011; Freckleton, 2011; 55 Grueber et al., 2011; Nakagawa and Freckleton, 2011; Richards e”
of construction probability on word durations during spontaneous incremental sentence production (2012, methodology): “Thus Burnham & Anderson (2002; 2004) and subsequent literature (e.g., Smith et al., 2009; Freckleton, 2011) warn against mechanistic initial selection and data dredging of variables that are not substantiated by either theoretical considerations or prior empirical research.”