The Report

Week 10Due 14 May 2027Worth 35%

Two thousand words, plus your record. You run the experiment you pre-registered in week 6, you report what happened, and you say what the result is and is not allowed to mean.

What to hand in#

Five parts, in this order, because the order is the argument: the question, then what you did, then what happened, and only then what it is allowed to mean. The word count covers the first four. The record is not counted, and it is not a summary: a record edited down to the lines that support the conclusion is the one part of this that a marker cannot check against anything.

1. The question, copied from your pre-registration without editing it. If you want to report a different question, that is a departure, and it goes in part 2 with the others.
2. What you did, including every departure from the pre-registration, each with the date you made it and whether you had seen any data by then.
3. What happened. The numbers, and for each one, where in the record it came from.
4. What you can conclude, and what you cannot. This is the part the marking is mostly about, and it is where the sentence from the week 9 lab belongs: where the day your experiment began sat in your own record, and how much of your result you therefore cannot separate from the day.
5. Your record, whole and raw. The actual log, with the gaps left as gaps.

The largest criterion is inference, not result#

Forty-five per cent of this piece is whether your conclusion follows from your data. Not whether your day improved. Not whether the intervention worked. Not whether you found anything at all.

A report that says “I measured this carefully for two weeks, and the honest answer is that I cannot tell, and here is precisely why” is a strong report. It may be the strongest kind. A report that says “this worked, I feel much better”, on a design that cannot support the claim, is a weak one no matter how true the feeling is.

This is not a stylistic preference. Marking the outcome would reward regression to the mean: most of you began your experiments on an unremarkable or bad stretch, and most of you will have improved regardless of what you did. A course that handed out marks for improvement would be handing them out for arithmetic, and week 9 exists to show you exactly that.

What each criterion is looking for#

Written out so that you can apply them to your own draft before a marker does, and so that two markers reading the same report arrive at the same place.

Whether your conclusion follows, 45 per cent#

The marker reads your conclusion first and your design second, and asks whether that design could have produced that sentence. Three rival explanations are standing in the room for every one of you, and the report is expected to name all three and either dispose of one or concede it: your own expectation, which week 7 showed you cannot blind away; something else that moved at the same time, which is week 8; and where your starting day sat in your own record, which is week 9. A conclusion stated at a strength the design supports, with the three named and handled, sits at the top of this criterion.

Marks come off for a claim the design cannot reach, for a conclusion that never refers to a number in the record, and for leaving a rival unnamed. They do not come off for a conclusion that is small, hedged, negative or uninteresting, provided the hedging is doing real work rather than covering an absence of thought. Nothing here is affected by whether the conclusion turns out to be correct: nobody marking this knows, and your one subject cannot settle it.

The limits you declare, 35 per cent#

The test is not how many limits you list. It is whether each one changes what the report is then allowed to say. A limit that appears in part 4 and leaves the conclusion exactly where it was has not been used, and a report whose limits section could be moved unchanged onto somebody else’s experiment has not declared any.

So: specific to this experiment, quantified wherever your record can quantify it, and visibly narrowing the conclusion. “This study has limitations” earns nothing. “I lost the Thursday and the Friday of the removal week, which leaves three days in that phase against seven in the others, so the comparison is between a week and a long weekend” earns a lot.

Departures are marked here too. Deviating from your pre-registration is normal and costs you nothing. An undeclared departure that the marker finds by reading your record against your pre-registration costs you this criterion, and it is the one thing in this course that reads as dishonest rather than inexperienced.

Whether a stranger could re-run this, 20 per cent#

Literal, and checked literally. The marker takes your record and asks whether they could produce your numbers from it without asking you a question. That needs the measure and its units written down, the time of day it was taken, and every number in the report traceable to a dated line. Missing days appear as gaps, because a record with the gaps quietly closed up is a record that cannot be checked.

Presentation is not marked. Neat handwriting, charts and a title page are worth nothing here, and a photograph of a legible paper log loses nothing.

Null results#

If nothing happened, report that nothing happened, at the same length and with the same care you would have given a positive result. A null result is not the absence of a finding, and the difference is where the marks are. “I found no difference” on its own says nothing, because a design that could not have detected a difference produces that sentence too. The version that scores is “I found no difference, and given how far my two instruments disagreed with each other in week 3, the smallest difference I could have detected was larger than the one I went looking for.”

The reason this is insisted on is not politeness about disappointing semesters. Week 12 carries the numbers on what happened to a literature that published only the trials that found something, and the personal version has the same shape. A semester of reports that all found something would tell us only that we had taught you to find things.

If the stopping rule fired#

Stopping early is a result and is reported as one. Say when it fired, what it fired on, and what the days you did collect can and cannot support, which will usually be less than you hoped and more than nothing. A report built on four days that is honest about being built on four days can score above a complete one that is not.