Removing abiotic v. biotic status from this full model, as well as any interactions including abiotic v. biotic status, marginally decreased the likelihood of the model, but AICc criteria, which penalizes for an increase in parameter number, showed strong support for the reduced model that excluded abiotic v. biotic status (Table 3a).
On every iteration, the Bayesian information criterion (BIC), which penalizes model complexity, is computed.
The number of knots used for the splines was determined by comparing the Akaike information criterion (AIC) model fit statistics for a range of choices for number of knots and choosing the value that minimized the AIC statistic, which penalizes models with more knots to avoid overfitting [ 6].
Because the models have the same number of parameters, there was no need to use more sophisticated information or theoretical criteria such as the Akaike's information criterion (AIC) [41], [42] or the improved versions of AIC which have been proposed and applied to Monte Carlo models by Bozdogan [43], [44], since the term which penalizes complex models would be redundant [45].
The semiparametric approach combines a parametric model with a different substitution rate on every branch with a nonparametric roughness penalty which penalizes the model if rates change too quickly from branch to branch.
QICC is the corrected version of the quasi-likelihood under the independence criterion, which penalizes for model complexity.
The variational method is based on the minimization of the continuous energy functional which penalizes all deviations from model assumptions: E ( u ) = ∫ Ω ( M ( D k f, u ) ⏟ Dataterm + μ S ( ∇ f, ∇ u ) ) + D ( div u ) ⏟ Regularizer dx, (3).
Model acceptability was quantified with Deviance Information Criterion DICC), which penalized model accuracy by the number of retrieved parameters.
We used the Akaike information criterion (AIC), which penalizes extra effective parameters to avoid overparameterized models, to select the minimum adequate model (Burnham & Anderson 2002).
Model selection was based on a fitness which penalized models for increasing complexity.
