informationCriterion: Information criterion and weights for a set of models

In maticce: Mapping Transitions in Continuous Character Evolution

Description

Returns AIC, AICc, and BIC values and weights for a set of models.

Usage

  informationCriterion(u = NULL, lnL = NULL, K, n = 1, names = NULL)
  ## S3 method for class 'hansenBatch'
informationCriterion(hansenBatch)
  ## S3 method for class 'informationCriterion'
print(x, ...)

Arguments

u	A vector of deviances, indexed by model.
lnL	A vector of log-likelihoods, indexed by model.
K	A vector of degrees-of-freedom / number of free parameters, indexed by model.
n	Sample size; for a phylogenetic comparative analysis, n = the number of tips.
names	Optional vector of model names, indexed by model.
hansenBatch	Output from runBatchHansen.
x	Output from informationCriterion.
...	Additional arguments to be passed along to print

Details

At the minimum, a vector of either the model log-likelihoods (lnL) or deviances (u = -2 * lnL) and a vector of number of free parameters for each model (K) must be provided for the function to work. If the sample size (n) is not provided, the function calculates AICc and BIC assuming n = 1. Information criterion statistics are calculated following Burnham and Anderson (2002).

Value

A list with the following vectors, all indexed by model number:

names	Model names; if not provided, a vector of 1:length(u).
u	Deviance.
K	Degrees of freedom.
AIC	Akaike information criterion.
AICc	Small-sample AIC.
BIC	Bayes information criterion.
AICwi	AIC weight.
AICcwi	AICc weight.
BICwi	BIC weight.

Author(s)

Andrew Hipp [e-mail]

References

Burnham, K. P., and D. R. Anderson (2002) Model selection and multimodel inference: a practical information-theoretic approach. Springer, New York.

maticce documentation built on May 2, 2019, 6:13 p.m.