1# snowcat731

You can calculate autocorrelation and partial autocorrelation and look up the shape of these two correlation to decide model structure (AR, ARMA, or ARIMA) and model order. As for detail, you can read book "Time Series Analysis Forecast and Control". This book provides you detail how to do it.

我有做过这个模型的,不过我都是用法语写的.我把我的基本程序放到下面,你照着试试看.

library(tseries)
sale=scan("C:\\Documents and Settings\\Dan\\BUREAU\\st\\xsales.txt")
ts.plot(sale)
shapiro.test(sale)
lsale=log(sale)
> ts.plot(lsale)
dlsale=diff(diff(lsale),12)
> ts.plot(dlsale)
par(mfrow=c(2,1))
> acf(dlsale)
> pacf(dlsale)
由上面的图我确定了SARIMA(1,1,1)*(1,1,0)12模型
然后再验证
m11=arima(x = dlsale, order = c(1, 1, 1), seasonal = list(order = c(1, 1, 0), period = 12))
对残差进行检验
tsdiag(m11)
r1=((m11$residuals)^2)
ts.plot(r1)
最后是GARCH模型
garch(lsale,order=c(1,0))

还有问题，可以给我邮件