r - auto.arima with allowmean = T is the same as allowmean = F -
im runing auto.arima forecast package allowmean = t , allowmean = f , exact same results - , both of them include intercept.
does has idea doing wrong , how resolve issue?
following repreducable example:
library(forecast) set.seed(1) z=arima.sim(n = 101, list(ar = c(0.8))) ts.plot(z) runing auto.arima allowmean = f produces:
> auto.arima(z,d = 0,max.p = 7,max.q=0,stepwise= f ,allowmean = f) series: z arima(1,0,0) non-zero mean coefficients: ar1 intercept 0.7275 0.4241 s.e. 0.0688 0.3012 sigma^2 estimated 0.7118: log likelihood=-126.52 aic=259.05 aicc=259.29 bic=266.89 which same allowmean = t:
> auto.arima(z,d = 0,max.p = 7,max.q=0,stepwise= f ,allowmean = t) series: z arima(1,0,0) non-zero mean coefficients: ar1 intercept 0.7275 0.4241 s.e. 0.0688 0.3012 sigma^2 estimated 0.7118: log likelihood=-126.52 aic=259.05 aicc=259.29 bic=266.89 i'll thankful ideas.
this appears bona fide bug. (note allowmean pretty recent addition - it's available in forecast 6.1, not in 5.1.) suggest contact package maintainer, rob hyndman, , point him question.
Comments
Post a Comment