Hi Kevin,
You should proceed with the ROC curve, I think. An example:
GET
FILE='C:\Program Files\SPSS\Coronary artery data.sav'.
ROC
time BY group (1)
/PLOT = CURVE(REFERENCE)
/PRINT = COORDINATES
/CRITERIA = CUTOFF(INCLUDE) TESTPOS(LARGE) DISTRIBUTION(FREE) CI(95)
/MISSING = EXCLUDE .
In the real cases, it is not possible to get both best sensitivity and
best specificity at the same time, but you are able to find a reasonable
compromise between the two given the ROC data. If you are able to figure
out the "prices" of errors of I and II types, and the prior probability
of the disease among tested persons, you can calculate the optimal
cut-off quite exactly.
Hope this helps
Jan
-----Original Message-----
From: SPSSX(r) Discussion [mailto:
[hidden email]] On Behalf Of
KEVIN MANNING
Sent: Friday, January 19, 2007 10:40 PM
To:
[hidden email]
Subject: Logistic Regression / Discriminant Analysis
Dear all,
Here is my dilemma:
I have 2 groups of participants, disease and no disease, and one
continuous variable.
I want to determine a cut-off score for the variable that produces the
best sensitivity and specificity. How can I do this?
I was under the impression this could be accomplished using logistic
regression, but I cannot figure it out (I would also like to view the
odds ratios that are produced as well). ROC curves were equally
confusing.
Any help at all would be most appreciated. Thanks!!!
Kevin
Locked
