Equality with threshold about buysetest

[bozenne]

Thanks, I have clarified that in the documentation:
-BuyseTest function: "A pair will be classified as neutral if the difference in endpoint is strictly below this threshold. "
-BuyseTest-package, summary function for BuyseTest have now some text describing the net benefit with formula (where the >= appears).
See new version on Github in the branch master.

The reason for using >= instead of > is that when using a threshold of 3 months, one usually expect that differences in survival of exactly 0 would count as a relevant difference while when using no threshold one usually expect that no difference does not count as relevant. This is why by default the threshold is set to a very small value (I think 10^{-12}).
This way the default setting does match table 2 of Buyse but it is true that when specifying a non-0 threshold there is a difference. We discussed that at some point with Marc Buyse and Julien Peron and we agree that was better like that.
So the software will stay like that.

[graemeleehickey]

Thanks! Appreciate the fast response. I'm using for continuous endpoints rather than event times. But your reasoning makes sense.