Date:         Thu, 4 Mar 1999 13:46:56 -0500
Reply-To:     Jeff Mather <Jmather@HARTHOSP.ORG>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         Jeff Mather <Jmather@HARTHOSP.ORG>
Subject:      sample size
Content-Type: text/plain; charset=US-ASCII

Hi,
I'm having trouble with a sample size problem and I'm hoping someone could give some hints.
The investigator is proposing a cross-over study for diagnosis of viability of different images of the heart.  He wants to compare diagnositic parameters (sensitivity, specificity, Pos. pred. value, Neg predictive value...) for both precedures against a gold standard for viable , non-viable.
 
He has given me the diagnostic parameters from a previous study for one of the procedures.
He is hypothesizing that the new procedure will have significantly greater diagnositic results.
 
I'm aware of methods to estimate sample size based on predicted sensitivity, specificity...etc with 95% c.i., but I'm not sure how to estimate sample size to compare these procedures such that the new procedure has adequate power to show a significantly greater sensitivity or specificity...
 
Thanks for any help.
 
Jeffrey F. Mather
Research Analyst
Research Administration
Hartford Hospital
email: jmather@harthosp.org
Phone: (860) 545-2093