Caffeine industrial time series
Hamilton and Watts (1978) state this series is produced from a cyclic industrial process with a period of 5.
data(Caffeine)
The format is: num [1:178] 0.429 0.443 0.451 0.455 0.44 0.433 0.423 0.412 0.411 0.426 ...
The dataset are from the paper by Hamilton and Watts (1978, Table 1). The series is used to illustrate how a multiplicative seasonal ARMA model may be identified using the partial autocorrelations. Chatfield (1979) argues that the inverse autocorrelations are more effective for model identification with this example.
Hamilton, David C. and Watts, Donald G. (1978). Interpreting Partial Autocorrelation Functions of Seasonal Time Series Models. Biometrika 65/1, 135-140.
Hamilton, David C. and Watts, Donald G. (1978). Interpreting Partial Autocorrelation Functions of Seasonal Time Series Models. Biometrika 65/1, 135-140.
Chatfield, C. (1979). Inverse Autocorrelations. Journal of the Royal Statistical Society. Series A (General) 142/3, 363–377.
#Example 1
sdfplot(Caffeine)
TimeSeriesPlot(Caffeine)
#
#Example 2
a<-numeric(3)
names(a)<-c("AIC", "BIC", paste(sep="","BIC(q=", paste(sep="",c(0.85),")")))
z<-Caffeine
lag.max <- ceiling(length(z)/4)
a[1]<-SelectModel(z, lag.max=lag.max, ARModel="AR", Best=1, Criterion="AIC")
a[2]<-SelectModel(z, lag.max=lag.max, ARModel="AR", Best=1, Criterion="BIC")
a[3]<-SelectModel(z, lag.max=lag.max, ARModel="AR", Best=1, Criterion="BICq", t=0.85)
aPlease choose more modern alternatives, such as Google Chrome or Mozilla Firefox.