How far you can get in a day
This route is built to be taken as far as it will go in one focused day. You will not finish the report; you will finish everything that is hard to do alone. Notice where the time goes: more than half the day is spent before you process a single number, because choosing and checking the data is the part that decides whether the rest of it can work at all.
A fieldwork investigation is paced by a trip, and the weeks either side of it do the rest. This route has no trip, so it can be taken a long way in one sitting. What it needs instead is a running order, because a day without one ends with everybody still browsing repositories at three o’clock.
On a single day the dataset comes before the question, which inverts what the fieldwork route tells you. The reconciliation is that the gallery is organised by environmental issue rather than by publisher, so choosing a dataset is still choosing an issue. What a one-day route cannot survive is discovering at two o’clock that the question you fell in love with has no numbers behind it.
The day, block by block
Treat the minutes as proportions rather than a timetable. A class that finds its datasets quickly should spend the winnings on block 6.
The six criteria, and the one rule that sinks most secondary investigations: an inquiry that only reviews other people's writing is not a repeatable method, and the method criterion caps at the bottom band however good the rest of it is. Then the four tests any dataset has to pass before you spend the day on it.
Open two or three from the gallery and apply the four tests to each. Not reading about them: opening them, counting the rows, working out what one row is, finding the years that are missing. At least one of them will not be what its own row count says it is.
Bringing your own is better than taking one from the gallery, and the gallery exists so that nobody spends the day failing to find one.
Steps 1 and 2, written against the dataset you now have rather than in the abstract. The question has to name the indicator, its units, and the spatial and temporal boundaries, because on this route those are things you chose rather than things you measured.
Step 3, and it is the same job on both routes: one real strategy somebody is actually applying, two positions traced to named sources, one tension between conflicting goals. Each gallery entry names candidate strategies so this block does not become a search.
Step 4, which on this route is where the marks are made. Write the protocol so a stranger could rebuild your exact file: repository, dataset, version, filters, date accessed. Then decide your inclusion and exclusion rule and write down what it removed, before you have seen whether the relationship is there.
This block is the ethical core of the route. Choosing which years or which places to keep once you can see the pattern is the secondary equivalent of throwing away the quadrats that disagreed with you.
Step 5. Clean it, derive something the raw file does not contain, chart it properly, and run one statistical test you can justify. Everything here sits outside the word count, which makes it the best-value hour of the day.
Steps 6 to 8 sketched rather than written. Your findings in bullets, the limitations you already know the data has because you met them this morning, and an honest list of what remains.
The write-up, and only the write-up. Everything that is hard to do alone happens on the day: choosing the data, checking its shape, fixing the selection rule, and running the analysis. Block 7 exists so that the list of what remains is written down while you can still remember why each item is on it.