[6444] in s-news-athena
Time series with NA's
daemon@ATHENA.MIT.EDU (Christian Keller)
Wed Feb 15 02:40:36 1995
From: Christian Keller <keller@stat.math.ethz.ch>
Date: Wed, 15 Feb 1995 08:23:28 +0100
To: s-news@utstat.toronto.edu
Cc: keller@stat.math.ethz.ch
I would like to fit an AR-process to time series with missing values using
the function arima.mle.
The only restriction in arima.mle is that there mustn't occur any NA's in
the first p+d values, where p and d are the orders of the autoregressive
and differencing operators.
Does anybody know how many successive missing values are allowded in the
interior of the serie that the algorithm still gives efficient results?
A second question:
Does anybody has a version or a modification of the function ar.gm
(or another function which provides robust autoregression parameter
estimates) which can handle missing values?
Thanks for any suggestion.
*********************************************************************
Christian Keller e-mail: keller@stat.math.ethz.ch
Seminar fuer Statistik www: http://www.stat.math.ethz.ch/
ETH-Zentrum, SOL F6 phone: +41 1 632 3504
CH-8092 Zurich (SWITZERLAND) fax: +41 1 632 1086
*********************************************************************
