Date:         Wed, 16 Dec 1998 19:55:27 -0500
Reply-To:     "Dr. Mingwei Li" <Mingwei@SURGMON.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         "Dr. Mingwei Li" <Mingwei@SURGMON.COM>
Subject:      Re: Sensitivity and specificity
Comments: To: Hassane ABIDI <hassanab@LYON-SUD.UNIV-LYON1.FR>
Content-Type: multipart/alternative;

how to unsubscribe spssx-l?
    
    
    -----Original Message-----
    From: Hassane ABIDI <hassanab@LYON-SUD.UNIV-LYON1.FR>
    Newsgroups: bit.listserv.spssx-l
    To: SPSSX-L@UGA.CC.UGA.EDU <SPSSX-L@UGA.CC.UGA.EDU>
    Date: Wednesday, December 16, 1998 3:23 AM
    Subject: Sensitivity and specificity
    
    
    Dear list, 
    Here is my problem: 
    
    n1=   all patients where the diagnostic test was positive was selected, 
    n2=   patients randomly selected from the population where the diagnostic test was negative, 
    
    According to the real status to the disease we obtain the fourfold table: 
      
    
                    Disease status 
    
             | Yes  |    No 
         ____|______|___________ 
         T+  |  a   |  b    | n1    (all cases) 
         ____|______|______ |____ 
         T-  |  c   |  d    | n2 
         ____|______|______ |____ 
    
    May I estimate the sensitivity by a/(a+c) and specificity by d/(b+d) ? 
    
    Or I must take account the prevalence of disease (p) in general population, the positive and negative predictive values (estimated by a/n1 and d/n2 respectively) ? 
    
    I have the formulae  of the predictive values as function of the prevalence, sensitivity and specificity (Fleiss J.L., 1981 from page 4). 
 
    May I use the inverse of these formulae to estimate sensitivity and specificity ? 
    
    Any suggestions are welcome, 
    
    Hassane. 
      
    
 
[text/html]