labquiet_amy said:

Say what you would expect to see if you were wrong, before you look.

Careful with the causal language. Two things moving together in a period when five things changed is not a mechanism, and stating it as one makes the thread less useful to somebody reading it later.

Reporting back.

Update, and it is the boring outcome: nothing was wrong. The measurement I was reading had moved less than its own week-to-week variation, which somebody said upthread and I did not believe.

labquiet_amy said:

Say what you would expect to see if you were wrong, before you look.

Saving this. It is the first explanation that did not require me to already understand it.

Adding the clinical framing, because it changes how the question reads. Order of operations matters more than any single choice here: establish a baseline, change one thing, wait long enough for it to express itself, then measure again under the same conditions.

Old thread, but it comes up first when you search for this, so it is worth updating. One practical note: write down what you did and when, before you need it. Reconstructing a timeline from memory three months later is how people end up unable to answer the one question that would have resolved it.

Correct me if the detail matters more than I have assumed.