s-news
[Top] [All Lists]

How to define a new corStruct class ?

To: s-news@lists.biostat.wustl.edu
Subject: How to define a new corStruct class ?
From: thomas.dumortier@netcourrier.com
Date: Mon, 17 Oct 2005 23:44:05 +0200
Dear All,

I would like to model the following covariance structure in Splus 6.1(Windows):

V*V        V*W*r      V*W*s     V*W*t
V*W*r    W*W        W*W*u    W*W*u*u
V*W*s    W*W*u     W*W       W*W*u
V*W*t    W*W*u*u  W*W*u    W*W

Where 
- V*V and W*W are the variance component, and 
- r,s,t,u are <1

This matrix corresponds to a model with 2-period; the first period with one 
time point; the second period with 2 time points with AR(1) covariance 
structure; the correlation between the time point of the first period and the 3 
time points of the second period is unstructured.

I think that this can't be modelled by means of the standard corrStruct class. 
Specifically, the S+ helps mentions "Users may define their own corStruct 
classes by specifying a constructor function, and at a minimum, methods for the 
functions corMatrix and coef. For examples of these functions, see the methods 
for the corSymm and corAR1 classes". Unfortunately, the S+ help does not give 
any example.

I was wondering if somebody could help me (by an example) to define a new 
corStruct class by specifying a constructor function, and methods for the 
functions corMatrix and coef ?

Thank you very much

Best regards
Tom



 
Vr    V

-------------------------------------------------------------
NetCourrier, votre bureau virtuel sur Internet : Mail, Agenda, Clubs, Toolbar...
Web/Wap : www.netcourrier.com
Téléphone/Fax : 08 92 69 00 21 (0,34 ? TTC/min)
Minitel: 3615 NETCOURRIER (0,16 ? TTC/min)


<Prev in Thread] Current Thread [Next in Thread>
  • How to define a new corStruct class ?, thomas . dumortier <=