On Wed, 11 Jul 2007, Joseph Coveney wrote:
Also, I'm curious as to what problems Celine had in resolving the results in
the Stata Technical Bulletin article for -equim- and -equip with formulas in
the Hwang and Morikawa review article (full text at
http://www.diahome.org/content/abstract/1999/d3341205.pdf ).
Joseph Coveney
Joseph, I had the problem with resolving results in two different sources, not Celine.
The difficulty was in taking output from the STB article for -equim- and plugging those numbers (mean, sd, n) into the Hwang & Morikawa formulas for
Z1 = (d + delta) / s_d
lower limit of the one-sided CI = ( 2 - Z_0.95 * s_d)
Z3 = d / s_d
where d = the difference between means, s_d = sqrt(2 * s^2 / n1), and s is the pooled SD.
I was not clear on what to pay attention to in the unfamiliar -equim- output. It would have been convenient to figure out a way to use parts of the output as input for determining non-inferiority and superiority of a non-standard method.
Z1 and the lower limit of CI are used in Hwang & Morikawa, to establish non-inferiority for a given delta. Z3 and the lower limit of CI are used to establish superiority.
To try to get clarity on any pieces of the -equim- output I could use, I tried matching up the STB results with results I calculated by hand using the Hwang & Morikawa formulas. I couldn't get them to match.
Another problem was that d (difference between means) in Hwang & Morikawa was not identified as the absolute value of the difference, but I think it should have been. I could be wrong about that.
This method is not what I would have chosen. My hunch is that introducing delta, which is not generated from the data, into equations with parameters that are generated by the data, creates problems. I could be wrong about that, too.
That's all I have.
HTH.
Cathy Antonakos
*
* 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/
