DendroSync (version 0.1.4)

mod.table: Function to calculate goodness-of-fit statistics for variance-covariance models

Description

The function provides a table to compare fitted variance-covariance (VCOV) mixed models by AIC, AICc, BIC and LogLik. The restricted log-likelihood (LogLik) statistics for different models can be compared by Chi-square test, while Akaike information criterion (AIC), corrected Akaike information criterion (AICc) and Bayesian information criterion (BIC) are in the smaller-is-better form.

Usage

mod.table(modelList)

Arguments

modelList

a list of variance-covariance (VCOV) mixed models of the type as produced by dendro.varcov.

Value

The function returns a data.frame with rows corresponding to the objects and columns containing the following components:

n

the number of observations used in the model fit.

df

the number of parameters in the fitted model.

AIC

Akaike's Information Criterion of the fitted model.

AICc

corrected Akaike's Information Criterion of the fitted model.

BIC

Bayesian Information Criterion of the fitted model.

LogLik

log-likelihood of the fitted model

Details

The function returns a table to compare the fitted variance-covariance (VCOV) mixed models to the same data based on information criteria. The smaller AIC, AICc or BIC, the better fit. Also, LogLik value is included. AICc is calculated according to the formula AIC + 2*npar*(nobs/(nobs-npar-2)), where npar represents the number of parameters and nobs the number of observations in the fitted model.

References

Hurvich, C.M. & Tsai, C.L. (1989). Regression and time series model selection in small samples. Biometrika 76: 297-307.

See Also

AIC, BIC, logLik

Examples

# NOT RUN {
## Compare homoscedastic variance-covariance models on Iberian Peninsula
 # conifer ring chronologies using taxonomic grouping criteria (i.e. Species).
 data(conifersIP)
 ModHmSp <- dendro.varcov(TRW ~ Code, varTime = "Year", varGroup = "Species", 
                          data = conifersIP, homoscedastic = TRUE)
 
 mod.table(ModHmSp)# a data.frame containing information criterion values

 ## Compare homoscedastic variance-covariance models on Iberian Peninsula conifers
 # ring chronologies using geographic criteria (ie. Region).
 ModHmGoe <- dendro.varcov(TRW ~ Code, varTime = "Year", varGroup = "Region", 
                           data = conifersIP, homoscedastic = TRUE)
 
 mod.table(ModHmGoe)

# }