aic_weights: Akaike weights for a set of models

In umx: Structural Equation Modeling and Twin Modeling in R

Akaike weights for a set of models

Description

aic_weights converts a vector of AIC values into Akaike weights: the conditional probability that each model is the best in the set, given the data and the candidate models (Wagenmakers and Farrell, 2004).

Usage

aic_weights(x)

Arguments

x	Numeric vector of AIC values, one per model. NA values are not dropped.

Details

Weights are computed from AIC differences relative to the best model in the set:

w_i = \frac{\exp(-\Delta_i/2)}{\sum_j \exp(-\Delta_j/2)}, \quad \Delta_i = \mathrm{AIC}_i - \min(\mathrm{AIC})

The best model has ⁠\Delta = 0⁠ and therefore the largest weight. Weights sum to 1. A model 2 AIC units worse than the best has a likelihood ratio of ⁠\exp(-1) \approx 0.37⁠ relative to it, before normalisation across the set. The result is a statement about this candidate set only, not about the absolute probability that a model is true. $x^2 + y^2 = z^2$ Replaces MuMIn::Weights().

Value

Numeric vector of weights, the same length as x, summing to 1.

References

Wagenmakers, E. J., and Farrell, S. (2004). AIC model selection using Akaike weights. Psychonomic Bulletin & Review, 11, 192-196. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.3758/BF03206482")}

See Also

umxWeightedAIC(), AIC()

Other Reporting Functions: plot.umx_GWAS(), umx, umxAPA(), umxFactorScores(), umxGetLatents(), umxGetManifests(), umxGetModel(), umxGetParameters(), umxParameters(), umx_aggregate(), umx_time()

Examples

aic_weights(c(100, 102, 110))
# best model, 2 AIC worse, 10 AIC worse

umx documentation built on Oct. 11, 2026, 5:07 p.m.