此页讨论了如何导出ARFIMA中的差异参数d。
太好了,但是如何应用原始时间序列以产生新的差异序列呢?
一个算法很好,所以我可以从概念上理解,但是如果它能在ARFIMA中自动处理,那也很好。
https://online.stat.psu.edu/stat510/book/export/html/674
varve = scan("varve.dat")
varve = ts(varve)
library(astsa)
install.packages("arfima")
library(arfima)
y = log(varve) - mean(log(varve)) # Center the logs
acf2(y) ## ACF and PACF of the data
## Estimate d:
varvefd = arfima(y)
summary(varvefd)
d = summary(varvefd)$coef[[1]][1] #d = 0.3727
d #prints the value of d to the screen
se.d = summary(varvefd)$coef[[1]][1,2] #se = 0.0273
se.d #prints the standard error of d to the screen
##Residuals
innov = resid(varvefd)
plot.ts(innov[[1]])
acf(innov[[1]])