Date:         Tue, 2 Jun 1998 18:33:47 +0200
Reply-To:     Ralf Herold <herold@CHARITE.DE>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Ralf Herold <herold@CHARITE.DE>
Subject:      ? conf interv lin predictor proc logistic
Content-Type: text/plain; charset="iso-8859-1"

Dear SAS-users,
could You please help: I am searching how to calculate the confidence
intervals of a linear predictor obtained by a proc logistic binomial
analysis with one explanatory variable: this is known to be
1-alpha/2 confidence limits = 1 / (1 + e^(-eta +/- z(alpha/2) * sigma
(eta)), and 
eta = a + b' * x.
Thus, I need to know how to calculate sigma (eta), which, according to the
documentation (6.11, p. 420), is (1, x) * covariance matrix (parameter
estimates) * (1, x')'. 
Can you help with a calculation procedure for the latter term? Thanks!
 
 
Dr. med. Ralf Herold
 
* herold@charite.de - PGP welcome - Pieper 38056
* Kliniken für Kinderheilkunde und Kinderchirurgie
* Universitaetsklinikum Charite Campus Virchow-Klinikum
* Medizinische Fakultaet der Humboldt-Universitaet Berlin
* Telefon: +49-30-45066032  * Fax:  +49-30-45066906
* Augustenburger Platz 1, D - 13353 Berlin * Mittelalle 8