[19351] in s-news-athena
[S] How reliable is arima.mle ?
daemon@ATHENA.MIT.EDU (Kin Cheung)
Wed Sep 29 15:34:35 1999
Message-Id: <78E833D774B4D21190A30008C78C717E433328@exch1-ma.alkermes.com>
From: Kin Cheung <kin_cheung@alkermes.com>
To: "'s-news@wubios.wustl.edu'" <s-news@wubios.wustl.edu>
Date: Wed, 29 Sep 1999 15:24:21 -0400
Mime-Version: 1.0
Content-Type: text/plain;
charset="iso-8859-1"
Dear S+ users,
I am trying to simulate an ARMA(1,1) process and then
compare the estimated AR and MA parameters produced by
'arima.mle'. The estimated parameters are way off the
therectical values. I do understand that the estimates are
subject to sample variabilities. But after many times of
testing, I begin to ask how reliable is 'arima.mle'
estimates ? or is it something that I did incorrectly ?
The code that I used is:
set.seed(10)
arma <- arima.sim(model=list(ar=c(ar=-0.9,ma=-0.9)))
arma.fit <- arima.mle(arma, model=list(order=c(1,0,1)))
arma.fit
> set.seed(10)
> arma <- arima.sim(model = list(ar = c(ar = -0.9, ma = -0.9)))
> arma.fit <- arima.mle(arma, model = list(order = c(1, 0, 1)))
> arma.fit
Call: arima.mle(x = arma, model = list(order = c(1, 0, 1)))
Method: Maximum Likelihood
Model : 1 0 1
Coefficients:
AR : -0.36859
MA : 0.82566
Variance-Covariance Matrix:
ar(1) ma(1)
ar(1) 0.010412031 0.002540766
ma(1) 0.002540766 0.003835001
Optimizer has converged
Convergence Type: relative function convergence
AIC: 418.52584
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news
