addAIC.fun: Calculate the AIC for all one-covariate additions to the current model

Description

This function fits all models that differ from the current model by adding a single covariate from those supplied, and calculates their AIC value. It selects the best covariate to be added to the model, according to the AIC.

Usage

addAIC.fun(mlePP, covariatesAdd, startAdd = NULL, modSim = FALSE,...)

Value

A list with the following components

AICadd: Vector of the AIC values obtained from adding to the current model each covariate in covariatesAdd.
posminAIC: An integer indicating the number of the column of covariatesAdd with the covariate leading to the minimum AIC.
namecov: Name of the covariate leading to the minimum AIC.
AICcurrent: AIC value of the current (initial) model.
newCoef: A (named) list with the initial value for the coefficient of the best covariate to be added. It is used in stepAICmle.fun.

Arguments

mlePP: A "mlePP"-class object; usually the output from fitPP.fun. It defines the current model. The fitted model cannot include fixed parameters.
covariatesAdd: Matrix of the potential covariates to be added to the model; each column must contain a covariate.
startAdd: Optional. The vector of initial values for the estimation algorithm of the coefficients of each potential covariate. If it is NULL, initial values equal to 0 are used. Remark that in contrast to argument start of fitPP.fun, startAdd is a numeric vector not a list.
modSim: Logical flag. If it is FALSE, information about the process is shown on the screen. For automatic selection processes, the option TRUE should be preferred.
...: Further arguments to pass to AIC, for example the constant k for the AIC calculation.

Details

The definition of AIC uses constant k=2, but a different value k can be passed as an additional argument. The best covariate to be added is the one which leads to the model with the lowest AIC value and it improves the current model if the new AIC is lower than the current one.

Examples

Run this code


data(BarTxTn)

BarEv<-POTevents.fun(T=BarTxTn$Tx,thres=318, 
	date=cbind(BarTxTn$ano,BarTxTn$mes,BarTxTn$dia))

#The initial model contains only the intercept
 mod1Bind<-fitPP.fun(covariates=NULL, posE=BarEv$Px, inddat=BarEv$inddat,
	tit='BAR  Intercept ', 	start=list(b0=1))
#the potential covariates
covB<-cbind(cos(2*pi*BarTxTn$dia/365), sin(2*pi*BarTxTn$dia/365), 
	BarTxTn$TTx,BarTxTn$Txm31,BarTxTn$Txm31**2)
dimnames(covB)<-list(NULL,c('cos','sin','TTx','Txm31', 'Txm31**2'))

aux<-addAIC.fun(mod1Bind, covariatesAdd=covB)