st: non-rejection confidence intervals for relative risk

[Brad De Groot <datadoc@livestockis.com>]

Hello Statalisters:

I am working on vaccine efficacy data to submit to the USDA Center 
for Veterinary Biologics (CVB). One of the criteria used to establish 
a label claim is the lower bound of the prevented fraction calculated 
from the complement of the interval estimate of the relative risk 
(risk ratio). The method required by the CVB is called the "two 
one-sided score test" (TOSST) in a recent memo 
(http://www.aphis.usda.gov/animal_health/vet_biologics/publications/notice_07_15.pdf) 

and attributed to

Agresti A, Min Y, 2001. On small-sample confidence intervals for 
parameters in discrete distribution.
Biometrics 57: 963-971.

This method is implemented in StatXact. However, in addition to not 
being in financial condition to purchase yet another piece of 
software, I'm not in mental condition to learn yet another piece of software.

Is anyone familiar with an implementation of this within Stata? I 
found Roger Newson's 2004-03-18 post describing the somersd package 
available on SCC. However, if I understand that discussion correctly, 
that is a different method based on tansformation, coverage interval 
calculation from a symetric probability distribution (normal?) and 
back-transformation to calculate the confidence interval. I believe 
the method required by the CVB is classified as a non-rejection 
interval, which I think I understand only in a very general sense.

I will be thrilled if anyone can direct me to a Stata implementation. 
I will also be grateful any other general advice regarding this problem.

Thanks,

Brad




Brad De Groot, D.V.M., Ph.D.
Livestock Veterinary Service
Livestock Information Services
43 Rebich Lane
Dillon MT 59725
(O) 406 683 6613
(C) 406 660 1511 
*
*   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/