Re: [S] beyond level 2

cristofer_price@abtassoc.com
Thu, 23 Apr 98 11:17:37 est



Depending on how the sites were selected, and maybe how many there are,
perhaps it would be appropriate to treat the sites as fixed effects. Does
the paper you cite talk about that alternative? If anybody knows of any
good references that deal with deciding whether to treat sites as random
or fixed in multilocation trials, I would be interested in seeing them.
(The SAS System for Mixed Model; Littel, Miliken, Stroup, Wolfinger, 1996,
has a brief discussion on the topic).

In your case, with sites fixed, you'd have a two level model and your
immediate problem of Splus only handling two-level models would go away.

Programs that I am aware of that can handle three levels: SAS proc mixed,
HLM, MLn.

- Cris Price
Abt Associates

______________________________ Reply Separator _________________________________
Subject: [S] beyond level 2
Author: "DR. L.Y. HIN" <z044106@mailserv.cuhk.edu.hk> at internet
Date: 4/23/98 10:31 AM

Dear S-plus user,

I have a problem and will be grateful to all who help.

I will be conducting a multicenter trial over n centers, recruiting mi
patients in each of the i centers (i=1,...,n). For each patient, I will
obtain serial measurements over time t.

Ideally, I presume I should analyze the data using a three-level
hierarchical modelling approach, setting level 1 as the center-specific
level, level 2 as the patient-specific level, while level 3 being the
time-specific data.

However, I can't get nlme() or lme() to do multilevel modelling beyond
level 2. I thus turned to Oswald, and it seems to be able to handle up to
level 2 only as well (if I am correct). I came across the paper Y. Lee and
J.A. Nelder 'hierarchical generalized linear models' J.R.Statist. Soc. B
1996;58(4):619-678. seems to be able to cope with this situation. i wonder
if anyone has written and S algorithm for the multilevel methods or that
described in this paper?


Thank you.


Lin
----------------------------------------------------------------------
Dr. L.Y.Hin
Department of Obstetrics and Gynaecology,
Prince of Wales Hospital,
Shatin,
Hong Kong.
Tel.: (852) 26322807
Fax : (852) 26360008
----------------------------------------------------------------------

-----------------------------------------------------------------------
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

-----------------------------------------------------------------------
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