Uninformative Parameters and Model Selection Using Akaike's Information Criterion

By T. Arnold. 2010

3,353 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Ecological Modeling

View Paper: https://www.semanticscholar.org/paper/7a0dd286f414a8185ad155a8ab843c1313bec220

How researchers see this paper

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

In Artificial Intelligence

In Ecological Modeling

Citations in the corpus

Intensified agricultural use of grasslands reduces growth and survival of precocial shorebird chicks (2013, methodology): “Because this model did not lower the AICc while adding one parameter, this is not a competing model (Arnold 2010).”
Anthropogenic Noise Alters Bat Activity Levels and Echolocation Calls (2015, methodology): “Eight-five percent confidence intervals are more consistent with an AIC approach than are 95% confidence intervals (Arnold, 2010).”
Some mechanisms underlying variation in vital rates of grizzly bears on a multiple use landscape (2015, methodology): “…would be independent of each other and each could influence reproductive or survival rates, I first used all possible models as candidates followed by summing Akaike variable weights across all models to estimate the relative importance of each (Burnham and Anderson 2002:167–169, Arnold 2010).”
Mesopredator spatial and temporal responses to large predators and human development in the Santa Cruz Mountains of California (2015, methodology): “We found that a radius of 500 m minimized model Akaike Information Criterion (AIC; Anderson and Burnham, 2002; Arnold, 2010) and therefore used that scale for our development covariates.”
Megafires: an emerging threat to old-forest species (2016, neutral): “Arnold TW. 2010.”
A brief introduction to mixed effects modelling and multi-model inference in ecology (2018, background): “This makes inference easier because the resulting top model set will likely contain fewer parameters, and certainly fewer spuriously ‘significant’ parameters (Burnham & Anderson 2002; Arnold 2010).”
Breach of the northern Rocky Mountain geoclimatic barrier: initiation of range expansion by the mountain pine beetle (2012, methodology): “Non-nested models can be compared by examining information criteria, such as the Akaike information criterion (AIC), with the lowest AIC value identifying the models of best fit for a given data set (Arnold, 2010).”
Range shifts in response to past and future climate change: Can climate velocities and species’ dispersal capabilities explain variation in mammalian range shifts? (2018, result): “…model based on AIC was the full model (a model containing all four predictor variables) but there were other models within two AIC units, the simplest model was chosen as ‘best’ as long as the deviance explained was similar to that of the full model (Arnold, 2010; Burnham & Anderson, 2001).”
Real-time reliability evaluation of two-phase Wiener degradation process (2017, methodology): “It has been used for model selection and ranking of models in a variety of modeling research (Arnold, 2010).”
A farewell to the sum of Akaike weights: The benefits of alternative metrics for variable importance estimations in model selection (2017, background): “Because full reality is un - known, an AIC value in itself carries no information about how close is any particular model to reality independent of others.”
Life-History Characteristics of Mule Deer: Effects of Nutrition in a Variable Environment (2014, methodology): “We retained variables from the previous stage of an analysis if the parameter estimate differed from zero (i.e., 90% CI did not overlap 0) or had an importance weight >0.50 (Arnold 2010, Monteith et al. 2011).”
Improving the early detection of alien wood‐boring beetles in ports and surrounding forests (2015, methodology): “From the set of plausible models, we omitted the models with uninformative parameters, that is models with DAICc below two but including only one additional parameter compared to the best model (Arnold 2010).”