公开问题
What counts as having reproduced a published result?
问题维基
提问的人可以创建维基，并指定维护者，把共同的进展整理在这里。
发布于 2026-08-25最近更新 2026-08-28932 次浏览
A genuine question about practice rather than about a calculation.
We re-ran an analysis from a paper using our own pipeline and got a number that agrees within the quoted uncertainty, but our central value sits about away and our uncertainty is 30% larger because we did not have their full covariance.
Is that a reproduction? I have seen the same situation described as "confirmed" in one paper and "in tension with" in another, by people I respect equally. I would like to know what other groups take the line to be, because it affects how we write ours up.
提问者 · 2026-08-25Sofia Lindqvist
2 个回答
采纳的回答排在最前
回答
The framing I have found useful: a reproduction is a claim about the procedure, not about the number.
If your pipeline implements what the paper describes and lands within the uncertainty, you have reproduced it, and the 30% is a fact about what they published rather than about your agreement. What you cannot say is that you have confirmed the central value to their precision, because you did not reach their precision.
Write both sentences. They are different claims and conflating them is where the disagreement you are describing comes from.
回答
Agreeing with the above and adding the practical part: publish the covariance you did have. Half the time the next group can close the gap from it, and the reason this question keeps coming up is that almost nobody does.
登录后可以回答。