Treat your data
A downloaded file, cleaned, with an inclusion rule already applied and written down. And ideally a decision about which statistical test you are going to run; if not, make it before you start processing.
- Raw data tabulated, with the full dataset in an appendix
- At least one quantity the file did not contain
- Your own figures, built from your own values
- A statistical result, correctly calculated. No interpretation.
Six marks, almost none of the evidence costs a word, and one rule that is specific to this route.
Everything that earns marks here sits outside the 3,000: the tables, the graphs, the calculations. What costs words is describing them, and describing them earns nothing. The rule that belongs to this route: the figures have to be yours. A repository that draws a beautiful chart for you is offering you the one thing you cannot use.
Data tables, raw and processed. Every graph and chart. Equations, formulae and calculations. Diagrams and photographs. Citations and the bibliography.
Spend freely. Make them excellent.
The sentence saying what each table shows. Why you chose that statistical test. Your null and alternative hypotheses. What you did about outliers, and why.
About 200 words in total. That is all you need.
One caution: appendices are not read, so the tables you want marked go in the body, and the full dataset goes in an appendix.
One exception: you may write a sentence of interpretation under a figure, or straight after a test result. It is credited in Criterion E (Analysis and conclusion), wherever it sits.
Everything below is how we suggest you actually do it.
Choosing the graph and the test
The graph type is a real decision, not a default. Comparing two groups needs a bar chart with error bars. A relationship between two continuous variables needs a scatter plot with a line of best fit. To show how spread out your values are, and which values sit outside the rest, use a box plot. The box covers the middle half of the data and the line across it is the median. Anything past the whiskers is an outlier you will have to explain. Picking the wrong one hides the very pattern you are trying to show.
One chart is worth knowing, because it solves a problem students often give up on. When two variables both change over time, a bubble chart puts all three on one figure: time along the bottom, one variable up the side, and the second as the area of each bubble. State the scaling you used: area, not radius. Scale by radius and the eye reads the difference as far larger than it is.
Where you can, run an inferential test (a test of whether your pattern could be due to chance), with a null and an alternative hypothesis.
The two hypotheses are one sentence each, written before you test. The null says there is no effect: "there is no difference between the two groups". The alternative says there is one. It does not predict which way unless you have a reason to. You are not trying to prove the alternative. You are checking whether the data makes the null hard to believe.
The convention is p < 0.05. p is the probability of getting a result like yours by chance alone, if the null were true. Below 0.05 the result counts as statistically significant and you reject the null. Above it, you do not have enough evidence against the null. That does not show the null is true. The words you use matter as much as the number: data supports, partially supports or does not support a hypothesis. It never proves one. A report that says "proves" claims something statistics cannot give you.
Say which error bars you plotted: standard deviation, standard error or a 95% confidence interval. They answer different questions, and a reader cannot tell which by looking. Error bars show spread. They are not a significance test: only the test settles it.
Error bars and trend lines are the two things that get drawn and then never mentioned again. Draw them here, interpret them in step 6.
One habit costs marks for no reason. Run the tests your question needs, and stop. Running three tests when one was needed does not look thorough. It looks as if you were not sure which one answered the question.
You do not need to reproduce the arithmetic, the formula, or the rank-by-rank calculation. Nobody is marking your ability to add up, and every line of it costs words you need elsewhere.
What has to be visible is the decision trail: which test, why that test for that data, the value of the statistic, n (sample size) or the degrees of freedom, the critical value or p, the significance level, and the verdict on the null hypothesis. The credit is for justifying the choice, because that is a judgement. The arithmetic is not.
Tables are free, so put the rank table or the observed-and-expected table in the body, where a reader can see the test was set up correctly.
A line of best fit says the relationship has a shape, so it needs justifying like any other claim. Straight is not the default: fit the shape that theory or the scatter itself suggests. Your graphing tool can fit a curve where the data curves.
Two lines can never be justified. One is a line through a scatter whose correlation is close to zero. The other is a line across categories that have no order: sites and species have no slope to find. Plain dot-to-dot joining is acceptable, and with only a few points it is often the honest choice.
If you fit a line, add R², the share of the variation the line explains. It is not the same as the correlation coefficient, and quoting one as the other is a real error.
The presentation marks nobody means to lose
Criterion D's first strand is communication. None of what follows is difficult, and examiners notice all of it.
A table of means with nothing showing the readings they came from. The marks need both: raw data is evidence you collected something, processing is evidence you did something to it. One without the other is half a Criterion D.
Never the chart they already made
Most of the repositories you will use offer ready-made charts. Our World in Data will draw you a beautiful line for any indicator; a cantonal portal will map its own monitoring stations; a UN agency will publish the figure from its own report. Every one of those is a trap.
The marks here are for what you did to the numbers.
Take the download, not the picture. Every repository worth using has a button giving the values behind the chart, and building your own figure from them is about five minutes' work.
A screenshot of the publisher's own graph, captioned and cited.
Correctly credited, and still no evidence of any work by you.
Your own chart, from the downloaded values, showing the subset you selected and a quantity you derived, with your own axes and units.
Same underlying data, entirely different piece of evidence.
A published figure can appear in your background, in step 1, as a source like any other: cited, credited, and doing the job of explaining the issue.
What it cannot do is appear in your data section as though it were your processing. If it is somebody else's figure, it is a reference, and it belongs with your other references.
The shark file arrives as a spreadsheet, not a picture: 23,118 rows, each one a kind of animal caught on one kind of gear at one beach in one month. Any chart you find about shark control was drawn to answer its author's question, not yours.
The study turns on one percentage per gear: white sharks as a share of all the sharks that gear caught, in four areas, 1996 to 2023. That percentage exists nowhere in the download, so it can exist in nobody else's chart either. Every figure in the report is built from the downloaded rows.
If the publisher's chart already showed your finding, your finding was theirs.
Look at the chart the publisher already made and ask whether it answers your question. If it does, your question needs sharpening; if it does not, you now know exactly what your own figure has to show that theirs does not.
Raw, then processed, then stop
Show the raw data in a table.
Then process it into something the file did not contain. A downloaded column is raw material however tidy it looks.
Raw: one row per kind of animal, per gear, per beach, per month, with how many were caught in NumberCaught, straight out of the download. The whole file is 23,118 rows, so the body of the report carries a structured sample and the full set goes to an appendix nobody will read.
One row is not one animal. In the study's subset, 591 of 4,299 rows record more than one, so counting rows would undercount. Everything is done by adding up NumberCaught instead.
Processed: first a helper column, P, labelling every row: =IF(J2="WHITE SHARK", "White shark", "Other shark"), where J is CommonName. Then four SUMIFS, one for each gear and kind, adding NumberCaught (column O) by Gear (column F) and the helper. =SUMIFS(O:O, F:F, "Net", P:P, "White shark") gives 73, and the other three cells give 3,223, 74 and 1,858. Then one division per gear: 73 of 3,296 is 2.2% for nets, 74 of 1,932 is 3.8% for drumlines. Neither percentage is anywhere in the file.
That pair of percentages is the whole investigation. Everything in the conclusion is a statement about it, and there is not one number in the report that could have been read straight off the original file.
Derive at least one quantity your download does not contain, and make it the one your conclusion turns on. It does not have to be clever: a share of each group's own total is a helper column, a few SUMIFS and a division, and it is the difference between treating data and tabulating it.
Gaps, zeros and ceilings
Downloaded data is missing for reasons nobody wrote down, and the file rarely says which.
Three things to look for, and to say what you did about. A gap: a year or a unit simply absent. Your chart must not quietly join the line across it. A zero: sometimes it really means none, often it is a value below the detection limit, and sometimes the compiler filled it in where nothing was reported at all. Those are three different claims written as the same digit. A ceiling: a value that appears again and again at exactly the top of the range. It is usually the reporting limit, not a measurement.
None of these has one right answer. All of them have one right behaviour: notice it, decide, and say so in a sentence. An unexplained gap is a hole in your method; an explained one is a limitation, which is worth marks in step 7.
The gap is at the end, which is the worst place for one. The file runs from 1996 to May 2026, so its last year is five months long. Plotted by year without checking, 2026 sits on the chart looking like any other year, and whatever it shows is five months set against twelve.
The zeros are never written down. A month in which no white shark was caught has no white-shark row at all, not a row reading 0: a missing row is a zero. Adding up NumberCaught gets that right without trying. Averaging the rows that exist quietly leaves out every month with nothing in it, and makes the catch per month look bigger than it was. Nor can an absent row tell you whether a beach caught nothing or had no gear in the water, because the file records catches, not effort.
And then the fourth thing, which the rows themselves never announce: the collector changed what it was doing. From 2024 new locations, daily servicing and new drumlines came in. All sharks caught ran at about 500 to 850 a year until 2022, then 948, 1,497 and 3,430 in 2023 to 2025. A catch that more than trebles in two years while the gear changes cannot be read as the sea changing, so the series was cut at 2023 and the reason stated.
Before you calculate anything, count the rows in each year, look hard at the first and the last, and ask what an absent row means. A thin final year, a zero that is never written down and a collector that changed its method are three things a downloaded file will not tell you about itself, and all three change what your totals mean.
Which test, and how to justify it
The statistics do not care where your numbers came from. Start from the claim you are trying to make, not from the name of a test.
Two periods, three catchments, countries above and below a threshold. You are asking whether the groups sit in different places.
As one rises, does the other rise or fall? Both measured on the same unit, so they come in pairs. This is where a second dataset joined on date or place is useful.
Counts falling into classes: animals by gear and species, stations by quality band, countries by income group. Tallies of things, not measurements of them.
One piece of vocabulary, because the table depends on it. Interval data is measured on a real scale with units. Ordinal data is ranked but not measured, so the gaps mean nothing: a five-point quality band, a stanine. Categorical data is counted into named classes. A downloaded file will happily give you all three in adjacent columns, and the column heading will not tell you which is which.
| Test | The question it answers | What it needs | Check before you use it |
|---|---|---|---|
| t-test | Do two groups differ, on average? | Interval data, around 10+ per group | Each group roughly normal, plot a histogram and look |
| Mann-Whitney U | Do two groups differ, when the data is skewed or ranked? | Interval or ordinal, 5+ per group | No normality needed; the two distributions should be a similar shape |
| ANOVA | Do three or more groups differ? | Interval data, around 10+ per group | Roughly normal; a significant result says only that some pair differs, so a post-hoc test such as Tukey is what names which one |
| Pearson's r | Do two measurements rise and fall together in a straight line? | Interval data, 10+ pairs, 30 is better | The scatter plot looks linear, and no single outlier is driving it |
| Spearman's rank | Do they move together, when the relationship bends? | Interval or ordinal, 7+ pairs | Consistently one direction; it tolerates curves and outliers that Pearson will not |
| Chi-squared | Are counts spread across categories differently from what you would expect? | Raw counts in categories | Every expected value is 5 or more, and they are counts, never percentages or densities |
Plot a histogram of each group before you assume normality. It takes a minute and it is the difference between a t-test you can justify and one you cannot. This is not the same as choosing the test after seeing the result: checking each group's own spread is part of choosing the test, but picking the test from the graph of the relationship you hoped to find is not. And decide whether your question is one-tailed or two-tailed before you run anything; most comparisons are two-tailed, and choosing one-tailed only to get a smaller p-value will be noticed.
One trap belongs to this route in particular. With a large downloaded dataset, significance is easy to get. Run a correlation on thousands of rows and almost anything comes out significant, including relationships far too weak to matter. Quote the strength as well (r or R squared), and talk about the size of the effect rather than how small p is.
The white shark comparison is an association question, not a difference one. Every shark caught falls into one of two gears and one of two kinds, white shark or other shark, so the data is a 2 by 2 table of counts: 73 and 3,223 for nets, 74 and 1,858 for drumlines. Counts in categories means a chi-squared test of independence, run on those counts and never on the percentages. Its condition is that every expected count is 5 or more. Here every one is above 50; the smallest is about 54.
The result: chi-squared = 11.6, df = 1, p about 0.0007, against a critical value of 3.84 at p = 0.05. The null hypothesis, that the kind of shark caught does not depend on the gear, is rejected. With 5,228 sharks in the table a small difference can clear that bar, so the report puts the size beside it: 2.2% against 3.8%, a gap of 1.6 percentage points.
Saying all of that takes one sentence: what was being asked, what kind of data it was, which condition was checked, and what the check found.
Name the test and its justification in the same sentence, and check the condition rather than assuming it. "Counts of sharks in a 2 by 2 table, with every expected count above 50, so a chi-squared test of independence" is one clause and it is worth a mark.
It can explain what a test does and what its conditions mean, and help you write the spreadsheet formula that turns your raw column into the one you need.
It cannot choose the test, run your numbers or tell you what a value is. A tool asked for a figure will produce one that looks right, and on this route a fabricated number is indistinguishable from a real one in the finished report. Every value comes from your file.
Ready for step 6?
Secondary data checklist0 of 18Four of these belong to this route, and building every figure yourself rather than copying the publisher's is the one that most often costs marks. The presentation checks at the end are a third of this criterion and the cheapest marks in the report.
Everything you have just been told not to do is what step 6 is for. The patterns you can see in your graphs, what is causing them, and what it means for your research question: all of that goes there, where six marks are waiting for it.
The white shark investigation used on the secondary-data route of this guide is the author’s own analysis of a published dataset: the Queensland Shark Control Program’s record of every animal caught on its nets and drumlines, published by the Queensland Government under CC BY 4.0 and downloaded on 26 September 2026. The choice of the four areas, the analysis and the conclusions are the author’s, not the Queensland Government’s.