ZeBook (version 0.4)

AICf: Calculation of AIC, Akaike's Information Criterion, given a vector of observation, a vector of prediction and number of parameter.

Description

This function calculate AIC criterion given a vector of observation, a vector of prediction and number of parameter. Note that number of parameters should include variance. AICcomplete is the same calculation of the AIC function of R (AICcomplete = n*log(RSS/n)+n+n*log(2*pi)+2*p, with p including variance). AICshort is the calculation described in chapter 6 Working with crop model (AICshort =n*log(RSS/n)+2*p, with p including variance). difference between AICcomplete and AICshort is AICcomplete-AICshort=n+n*log(2*pi) As you use AIC to compare models (with different number of parameters) on a same data (with same n, number of observation), you can use AICshort or AICcomplete.

Usage

AICf(Yobs, Ypred, npar)

Arguments

Yobs

: observed values

Ypred

: prediction values from the model

npar

: number of parameters (should include variance that count for one supplementary parameter)

Value

• a vector with AICcomplete and AICshort

Examples

x=c(1,2,3,4,5)
y=c(1.2,1.8,3.5,4.3,5.5)
fit = lm(y~x)
AIC(fit)
AICf(y,predict(fit),3) # 3 parameters : intercept, slope and variance