What I tell every new graduate student in week one

Every autumn I have the same conversation, so I have started writing it down.

Write the hypothesis first

Before a single sample is prepared, write one sentence: if X is true, then I expect to see Y. If you cannot finish that sentence, you are not ready to run the experiment — you are ready to run a pilot, which is a different thing and should be labelled as such.

The discipline here is not bureaucratic. A hypothesis you wrote down last week cannot be quietly reshaped by this week's results.

Keep a record you would show a stranger

Your notebook is not for you. It is for:

1. You, eighteen months from now, when a reviewer asks what buffer you used.
2. Whoever inherits the project.
3. Anyone who needs to check the work.

Say it plainly when it doesn't work

The single most useful sentence in science is "the data don't support the hypothesis." Practise saying it early, while the stakes are small.

Result: no difference between groups (p = 0.41, n = 12).
Interpretation: the effect, if present, is smaller than we powered for.

That is a finding. Write it up.