Usually, if one wants to compare two diagnostic tests against a gold
standard, one is interested in both the false positive and false negative
rates. One of the methods for comparing diagnostic tests with a gold
standard is ROC analysis (see -help roc-). ROC analysis helps one to find
the optimum balance of sensitivity and specificity. However, according to
the information you've provided, your friend's study appears to only consist
of true positives, so apparently he is not interested in the false positive
rate or the specificity. If this is so, it will not be amenable to ROC
analysis. If he is truly interested only in the false negatives, then all
you really need to do is compare the proportion of false negatives in test A
with the proportion of false negatives in test B. You can play around
recoding the Low, Medium, and High ratings and compare the proportion of
false negatives among the various recodes. With only 15 cases, and no
apparent concern for false positives, and with no other information about
the diagnostic tests other than you've given, it should not be too difficult
to eye-ball which recode would yield zero false negatives, in which case you
really don't need statistical analysis, right?.... :-) I'm kidding, of
course, but I say it to point out that there isn't much point in doing a
statistical analysis if you only have true positives in the study and you
can recode the ratings anyway you want.
Kidding aside, my guess is that your friend is interested in the optimum
balance between sensitivity and specificity, and that he has some non-cases
in the study. ROC analysis is probably what you want here, and the manual
has a good explanation and references. There is also -diagt- on SSC.
Regards,
Lee
Lee Sieswerda, Epidemiologist
Thunder Bay District Health Unit
999 Balmoral Street
Thunder Bay, Ontario
Canada P7B 6E7
Tel: +1 (807) 625-5957
Fax: +1 (807) 623-2369
[email protected]
www.tbdhu.com
> -----Original Message-----
> From: [email protected] [SMTP:[email protected]]
> Sent: Tuesday, February 18, 2003 11:41 AM
> To: [email protected]
> Subject: st: False negatives in clinical work
>
> I'm trying to help a pathologist friend with a study and am not sure how
> to
> proceed. She has 15 cases that are positive for a disease. There are two
>
> standard methods for diagnosing the disease from tissue samples, say A and
> B.
> A and B each produce a rating of High, Medium or Low on the indicator
> used
> for diagnosis. Each case has been rated by each method. She would like
> to
> know if there is a statistically significant difference in the number of
> false negatives between the two methods. It would also be useful to know
> if
> collapsing, for example, High and Medium into one category yields a lower
> false negative rate.
> Any suggestions on how to proceed most welcome.
> Best,
> Tom Dietz
>
>
>
> Professor of Sociology and Crop and Soil Sciences
> Acting Director of the Environmental Science and Policy Program
> Acting Associate Dean for Environmental Science and Policy, College of
> Agriculture and Natural Resources, College of Natural Science, College of
> Social Science
> Michigan State University
> Off campus phone: 802-372-4389
> Off campus fax: 802-372-4389
> Email: [email protected]
> Office: 118 Linton
> On campus phone: 517-355-0302
> On campus fax: 517-353-3355
> *
> * For searches and help try:
> * http://www.stata.com/support/faqs/res/findit.html
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
