[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
Re: st: Sample size determination for ROC comparison
At 00:43 15/04/04 +0900, Joseph Coveney wrote (in reply to Gregory Daniel):
I think that this can be done in Stata using -simulate-. Samples of various
Another useful tool might be the -powercal- package, downloadable from SSC.
If you can estimate a standard error for a known sample size using
-simulate-, then you can use that standard error for input to power
calculations using -powercal-, as described in the -powercal- manual
(distributed with the -powercal- package).
sizes with given areas under the reciever operating characteristic curves
be created with -drawnorm outcome predictor, corr(1 `rho' \ `rho' 1)
n(`sample_size')-. Then -replace outcome = outcome > 0.5-, and -replace- the
predictor according to the assumption for its distribution. (For a normally
distributed predictor, AUC given by -lroc- after -probit outcome
(`rho' + 1) / 2). Create two such samples (one for each of the two
sample comparison groups) with their given AUCs determined by their
`rho' values used in -drawnorm-, and -append- them, using a grouping
variable to tell them apart and to reference in the -by- option in
Power to discriminated the two groups at each sample size and level of Type I
error rate can then be estimated based upon the results returned from
-roccomp-, with the number of replications determined by the precision
for the power estimate.
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
Tel: 020 7848 6648 International +44 20 7848 6648
Fax: 020 7848 6620 International +44 20 7848 6620
or 020 7848 6605 International +44 20 7848 6605
Opinions expressed are those of the author, not the institution.
* For searches and help try: