Non-inferiority analysis

[Luca Cumi]

Hi!
I want to make a non-inferiority analysis, investigating if experimental group is not worse than control group. In my trial the percentage success in experimental group is 94%, and the percentage success in experimental group is 91%. The non inferiority margin is 15%.

Code:
ttest successyesornot, by (groups)

Is it correct the code? I have the difference and 95% confidence interval.

How do I calculate the p-value for non-inferiority?

Thank you for your time

[Joseph Coveney]

. . . the percentage success . . . and the percentage success . . .

Code:
ttest successyesornot, by (groups)

Is it correct the code? I have the difference and 95% confidence interval.

You probably will have a tough time convincing your colleagues that a Student's t test is best for comparing two proportions of success. Stata has a couple of official commands for the difference between two independent proportions and its confidence interval. There is also a user-written command for that, rdci, available from SSC.

How do I calculate the p-value for non-inferiority?

You could fit a generalized linear model for the risk difference using glm, but such a model sometimes will have convergence difficulties. You could also invert the confidence interval using, say, rdci in conjunction with minbound.