Posting this under the heading it deserves: Non-inferiority margins: how they are chosen and how they are abused Everything below is what sits behind that.
A question about non-inferiority margins that I think has a definite answer, unlike most of what I ask here.
I have the reasoning below and I am fairly confident about the direction. I am not confident about the size, and the size is what the decision turns on.
Trying to state the non-inferiority margins position in a way that someone who disagrees would recognise as fair, because I do not think the version in this thread passes that test.
Post #2 put the caveat in the right place and I want to underline it.
Absolute numbers, not just relative: a 30% relative reduction tells you the ratio but not the practical magnitude. The event rate in each arm and the difference between them tells you how many people benefit.
It is worth checking rather than assuming, which costs nothing.
Small methodological point on non-inferiority margins: repeating a measurement is cheap and resolves most of what is being argued about here at no cost to anyone.
Non-inferiority margins are chosen, and the choice is an argument rather than a fact. A wide margin can make a worse treatment look acceptable.
Adding a source would improve this post and I do not have one to hand.
Taking post #4 at face value and following it one step further.
I would keep non-inferiority margins and the decision it usually gets used for separate in this thread. They are related and they are not the same question, and merging them is why the last one went badly.
Quietly grateful for the plain phrasing. Not every thread gets that.
I read post #6 twice before replying, because I had assumed the opposite.
A treatment-policy estimand asks what happens to people assigned to a strategy, including those who abandon it. A hypothetical estimand asks what would have happened had everyone continued. Both are legitimate and they give different numbers.
It is one reading of the data and not the only reasonable one.
On post #9 — agreed on the reasoning, with one qualification.
Non-inferiority margins has a well-known answer and a correct answer, and the interesting work is establishing that they are the same. Nobody has done that here yet.
The documentation on non-inferiority margins is better than this thread and I say that as someone who has posted in the thread.
Following, with nothing to contribute beyond having asked the same thing elsewhere.
Post #12 is right about the mechanism and I think understates the practical bit.
Absolute and relative effects answer different questions. Write down the event rate in each arm and the difference between them; everything quotable is derived from those two numbers.
The mechanism is plausible, which is not the same as established.
Non-inferiority margins are chosen, and the choice is an argument rather than a fact. A wide margin can make a worse treatment look acceptable.
Risk of bias: structured appraisal of internal validity. Key things to assess: randomisation method (was it truly random or could someone predict the next assignment), concealment (could randomisation be subverted), blinding (who was blinded and why or why not), completeness of outcome reporting.
A single observation, in a thread that deserves better than single observations.
I had written a reply contradicting post #14 and deleted it. Here is what survived.
Reframing non-inferiority margins slightly, because I think the disagreement is about the question rather than the answer. If the question is "does it happen", yes. If it is "how often", nobody here knows.
Confirming post #17 from a second method, which matters more than confirming it from a second person.
Something worth flagging about non-inferiority margins: the strongest-sounding claims in this thread are the ones with no source attached, which is the usual pattern and not a coincidence.
A treatment-policy estimand asks what happens to people assigned to a strategy, including those who abandon it. A hypothetical estimand asks what would have happened had everyone continued. Both are legitimate and they give different numbers.
Anyone who has looked at this more carefully, please correct the record.
The arithmetic in post #17 is right; the assumption feeding it is the part to check.
Where the non-inferiority margins discussion usually stalls is that nobody wants to say "I do not know" and everyone is willing to say "it varies". Those are the same sentence with different clothes on.
Absolute and relative effects answer different questions. Write down the event rate in each arm and the difference between them; everything quotable is derived from those two numbers.
Stating my assumptions rather than smuggling them in.
Absolute numbers, not just relative: a 30% relative reduction tells you the ratio but not the practical magnitude. The event rate in each arm and the difference between them tells you how many people benefit.
Worth saying I have only my own numbers here, and n is small.
A treatment-policy estimand asks what happens to people assigned to a strategy, including those who abandon it. A hypothetical estimand asks what would have happened had everyone continued. Both are legitimate and they give different numbers.
The answer changed when I changed how I was measuring, which was informative.
Where I part company with post #24, and it is a narrow parting.
Non-inferiority margins is worth one more sentence than it usually gets, and the sentence is the one about how the number was arrived at.
Risk of bias: structured appraisal of internal validity. Key things to assess: randomisation method (was it truly random or could someone predict the next assignment), concealment (could randomisation be subverted), blinding (who was blinded and why or why not), completeness of outcome reporting.
The confident version of this sentence would be wrong, so here is the hedged one.
Adding a null result on non-inferiority margins. I looked, carefully, and found nothing, and null results deserve posting precisely because they never are.