s-news
[Top] [All Lists]

Re: AR for short, not equally spaced TS

To: Eric Zivot <ezivot@u.washington.edu>
Subject: Re: AR for short, not equally spaced TS
From: Spencer Graves <spencer.graves@pdf.com>
Date: Wed, 22 Jun 2005 11:00:02 -0700
Cc: "'Xao Ping'" <xao_ping@yahoo.com>, "'s-plus user list'" <s-news@lists.biostat.wustl.edu>
In-reply-to: <200506221740.j5MHepX3020642@mailhost1.u.washington.edu>
References: <200506221740.j5MHepX3020642@mailhost1.u.washington.edu>
User-agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.7.2) Gecko/20040804 Netscape/7.2 (ax)
ERIC: Does S-Plus have Kalman filtering software that would work with 6 short irregular time series?

Xao Ping: What do your 36 numbers look like? For example, do you get crudely a straight line with "qqnorm"? If yes, I think "lme" with "correlation=corCAR1" might give you something useful. I haven't used "corCAR1", but it looks like that should work. Of course, it would help to consult Pinheiro and Bates (2000) Mixed-Effects Models in S and S-Plus (Springer, esp. sec. 5.3). For testing, I think I would want to use "simulate.lme" as described in Pinheiro and Bates (sec. 2.4), in part because some of the things you might want to test would have parameters at a boundary, which would create problems with the usual asymptotics.

          spencer graves

Eric Zivot wrote:

One way to think of the irregular spacing is in terms of missing data. The true process works on equally spaced data but you ony observe data on specific dates. An AR type model can be fit using Kalman Filter techniques that automatically take care of the missing data.
------------------------------------------------------------------------
From: s-news-owner@lists.biostat.wustl.edu [mailto:s-news-owner@lists.biostat.wustl.edu] On Behalf Of Xao Ping
Sent: Wednesday, June 22, 2005 10:20 AM
To: s-plus user list
Subject: [S] AR for short, not equally spaced TS

Dear All:
I am analyzing a time course microarray experiment. I have only 6 chips taken in the moments of time 10, 501,2, , 85, 100 (relative units). I am thinking to apply time series analysis methods, such as AR or ARMA or ARIMA. I have several questions 1. Is it appropriate at all to think about autoregressive methods when the TS are not equally spaced? If the answer is "NO", then are there any alternative approaches? 2. Is it possible to extract any useful information having only 6 data points? Any relevant experience or advice are highly appreciated thank you in advance
Xao Ping
R&R Pharmakinetics Taiwan

__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
http://mail.yahoo.com


--
Spencer Graves, PhD
Senior Development Engineer
PDF Solutions, Inc.
333 West San Carlos Street Suite 700
San Jose, CA 95110, USA

spencer.graves@pdf.com
www.pdf.com <http://www.pdf.com>
Tel:  408-938-4420
Fax: 408-280-7915

<Prev in Thread] Current Thread [Next in Thread>