Experiment — ⟨EN: short name⟩

Project ⟨Pnn⟩ · notebook entry ⟨NNN⟩ · ⟨date⟩

A lightweight companion to notebook.md for a single sweep. Use this when an experiment is one variable and one afternoon; use the full notebook when it carries a hypothesis.

Question

⟨One sentence. What am I trying to find out?⟩

Independent variable

Held fixed

⟨Everything else. Being explicit here is what makes it a controlled experiment.⟩

Prediction

⟨Written before running. Specific enough to be wrong.⟩

Shape I expect: ⟨linear / concave / a knee at X / no effect⟩

What would surprise me: ⟨...⟩

Setup

hardware  :
software  :
data      :
seed      :
trials    :
command   :

Measurement resolution: ⟨the smallest effect this experiment can distinguish from noise. If you are measuring a rate p, you need ~100/p trials for a 10% relative standard error.⟩

Results

Verdict

Why

⟨The mechanism. If the prediction was wrong, what specifically was wrong in my model? Name a decision or a quantity, not a category.⟩

Secondary metric that explains it

⟨Which additional measurement closes the argument? If there isn't one, that is a finding: you are measuring the outcome but not the mechanism.⟩

What this changes

⟨A decision in the project, or a prediction for a later one. If nothing changes, say so — an experiment that changes nothing was worth running only if it ruled something out, and you should name what.⟩

Raw data

file    :
commit  :