Document resource
The challenge of knowing if something really is better, or if it is not worse, or if it is actually the same, is one that taxes statisticians and confuses many clinicians. And there are good reasons for that. All of these rely on us having an agreed measure of ‘betterness’. This can be easy in oncology, where it is still measurable in survival in many cancers. It can be trickier in other things, like moderate asthma control. With the measure locked in, we can then say if we are looking for new drug to beat an old drug ‘is it beyond reasonable to assume the two populations in this trial end up with the same value of ‘measure’ ?’ (And typically we want a tiny p value for this.)