Stephen Brealey writes:
>How would one calculate the necessary sample size to
>demonstrate the mean diagnostic accuracy between two
>groups of observers in comparison with a "gold" standard
>as being statistically significant?
You want to compare the sensitivity of one group of observers to the
sensitivity of a second group of observers. Likewise for specificity.
Is that right? Both sensitivity and specificity are proportions. So select a
sample size so that the test of two proportions has adequate power.
Alternately select a sample size so that the confidence interval for the
difference between two proportions is sufficiently narrow.
Any introductory Statistics text book should tell you the proper formulas.
One of my favorites is Norman and Streiner. Rosner is also a good choice.
Norman, Geoffrey R. and Streiner, David L. (1994) Biostatistics. The Bare
Essentials. St. Louis MO: Mosby-Year Book, Inc. (ISBN: 1-55664-369-1)
For the beginning student. A very readable book with a lot of humor.
Rosner, Bernard (1990) Fundamentals of Biostatistics, Third Edition. Belmont
CA: Duxbury Press (ISBN: 0-534-91973-1).
For the beginning student. A wonderful resource book for basic statistical
methods. I especially like the numerous examples of power and sample size
determination. Not quite as readable as some other Statistics books, though.
Steve Simon, [log in to unmask], Standard Disclaimer.
Professor Mean's Library: http://www.cmh.edu/stats/plibrary.htm
<http://www.cmh.edu/stats/plibrary.htm>
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|