The Chromozone / IAL Biology / Unit 1 / Topic 1

Directed study: exam technique

Exam technique: judging whether data support a conclusion

Spec 1.15, 1.16Directed study: exam techniqueExam technique: Data and conclusionsAbout 40 minutes
Back to the lesson

Directed study: exam technique. About 40 minutes, independent study, 23 marks. One technique: quoting figures from data and judging a conclusion, with points for and against, the difference between correlation and cause, and a final judgement. Type your answers in the boxes, then hand them in at the bottom of the page. Steps 1 and 2 stay open to use; the mark schemes open after you hand in: then mark your work point by point and correct it in a different colour. Teachers: set this after the lesson Health risk studies: correlation, causation and study design.

Teach these first

This page builds on this lesson. Teach it before students start here:

How this page works

One exam technique, in three steps: learn how to do it, see it done with the marks shown, then do it yourself on this lesson's content.

Step 1: how to do it 6 min

Some questions ask whether data support a conclusion. They test AO3: you use the data, then judge the conclusion. Follow these steps:

  1. Read the statement. Decide what the data would have to show for it to be true.
  2. Describe the overall pattern first. Then compare two values from the data. If there is no pattern, say so.
  3. Work the figures. Give a difference, a ratio or a percentage difference: (difference ÷ starting value) × 100.
  4. Points for the conclusion: a clear pattern, a large sample, a long study, other variables controlled.
  5. Points against: a small or biased sample, variables not controlled, a short study, overlapping error bars.
  6. Say "correlation", then give another explanation: a confounding variable, chance, or the cause the other way round.
  7. Finish with a judgement: "supports", "partly supports" or "does not support", with the main reason.

How Pearson rewards it. Each point earns its mark only if it uses the data or the study. A number copied from the table with no comparison often earns nothing. In a 6-mark "to what extent" question, marking goes in levels: describing the data (Level 1), then limitations of the data (Level 2), then what each limitation means for the conclusion (Level 3). So never stop at "the sample is small": say what that does to the result.

What earns a mark and what does not
You writeMark?
"Cases rise as the dose goes up, from 4 to 11 per 1 000 people: almost 3 times as many."Yes: the overall pattern, then a comparison worked out from the figures, with the unit.
"More people get ill."No: no pattern and no figures from the data.
"The error bars overlap, so the difference may not be significant."Yes: it says what the overlap means.
"The error bars show the results are reliable."No: examiners want the link to significance (do the bars overlap?), not to reliability.
"The groups differ in mean age, so age could explain the gap."Yes: a confounding variable, named and linked to the result.
"The study is not valid."No: say what was not controlled and what that does to the result.

Common misconceptions and exam errors

These come up again and again in the reports on real exam answers. Each one costs marks.

  1. Describing single values instead of the overall pattern. Examiners find it hard to give marks when answers list each bar or row. Start with the overall pattern or the conclusion, then add the figures.
  2. Wording that says one thing causes the other. Examiners saw definitions and descriptions that said a change in one factor "leads to" a change in the other. Write "linked to" or "goes with". When the question gives 2 or more marks, also say that the data do not show a cause.
  3. Error bars: calling them reliability, or naming them without a meaning. Link error bars to whether a difference is significant, not to reliability. Do not stop at "there are error bars": say what the overlap, or the lack of error bars, means for the conclusion.
  4. Controlled variables that are too vague. Write "the amount of saturated fat in the diet", not "diet", and "hours of exercise per week", not "exercise". Then say why it matters: it also changes the risk.
  5. Thinking the two groups must be the same size. Rates per 1 000 people can be compared when the groups differ in size. The faults to name are a group that is too small, and groups that differ in age or smoking.

Step 2: a worked example 6 min

Question: A cohort study followed 11 000 adults for 8 years. Each person said how long they sat each day. The table shows the results.

Daily sitting time and deaths from CVD in 8 years (invented data)
Sitting time / hours per dayNumber of peopleMean age / yearsDeaths from CVD per 1 000 people
Under 42 100416
4 to 85 600479
Over 83 3005615

Comment on how far the data support the conclusion that sitting for over 8 hours per day causes death from CVD. (4)

Answer 1 2 of 4 marks

The data support the conclusion. The more hours people sit, the more people die of heart disease (1). This proves that sitting causes heart disease. The people who sat for over 8 hours were older (1).

Why: The pattern earns one mark, but it has no figures. "Proves" claims a cause, which data from a study like this cannot show, so that sentence earns nothing. The age point earns one mark. It is not linked to the deaths, so it does not score a second mark. The judgement ("support") ignores the age point the answer has just made.

Answer 2 4 of 4 marks

There is a correlation: the death rate rises as sitting time rises (1), from 6 to 15 deaths per 1 000 people, which is 9 more per 1 000 (1). But the mean age also rises, from 41 to 56 years (1). Older people are more likely to die of CVD, so age, not sitting, could explain the rise: a confounding variable. The data do not show a cause. They only partly support the conclusion (1).

Why: Pattern with a correlation (1), a worked comparison with the unit (1), the age data used (1), and the link to age as a confounding variable plus a judgement (1). The large sample (11 000 people), the 8 years of the study, smoking, diet and the lack of error bars are also creditable points, but the answer is already at 4 marks.

Step 3: your turn 23 min

Seven questions on health risk studies. Use the steps from step 1. The data are invented.

  1. Questions 1 to 5 use this study. 8 400 adults aged 40 to 65 joined a free fitness app. The app recorded how long each person walked every day. The researchers followed them for 3 years and counted the new cases of CVD. Smoking and diet were not recorded.

    Walking time and relative risk of CVD in 3 years (invented data)
    Walking / minutes per dayRelative risk of CVDError bar: range
    Under 10 (baseline)1.00none
    10 to 290.880.80 to 0.97
    30 to 590.760.68 to 0.85
    60 or more0.720.62 to 0.84

    Describe the relationship between walking time and the relative risk of CVD. Use the figures, and use the frame. (1.15) (3)

    Overall pattern
    A comparison worked out from the figures
    How the pattern changes at the top end
  2. Calculate the percentage by which the relative risk of the group walking 60 or more minutes per day is lower than that of the group walking 10 to 29 minutes per day. Give your answer to 1 decimal place. (1.15) (2)

    Use the table in question 1.

  3. A newspaper headline says: "Walking for 30 minutes or more a day halves your risk of CVD." Use the data to explain how far the headline is supported. (1.15) (3)

    Use the table in question 1.

  4. A student says: "Drinking more tea raises the risk of CVD." Use the data to say whether this is supported. (1.15) (3)

    Tea drinking and relative risk of CVD (invented data)
    Cups of tea per dayRelative risk of CVDError bar: range
    None (baseline)1.00none
    1 to 20.980.90 to 1.07
    3 to 41.020.93 to 1.12
    5 or more0.990.87 to 1.13
  5. Suggest two reasons why the researchers cannot conclude from the walking study that walking more reduces the risk of CVD. Use information from the study. (1.15, 1.16) (4)

    Use the study described before question 1.

  6. Questions 6 and 7 use this study. Researchers followed workers in one large company for 10 years. They recorded each person's job and whether they smoked. Diet and exercise were not recorded.

    Shift work and CVD in 10 years (invented data)
    Day workersNight-shift workers
    Number of people4 8001 200
    Mean age / years4544
    Smokers / %2138
    New cases of CVD per 1 000 people1827

    Suggest how the researchers could use their data to test whether smoking, and not night-shift work, explains the difference in CVD. (1.16) (2)

  7. A union leader says: "Working night shifts causes CVD." To what extent do the data support this statement? (1.15, 1.16) (6)

    Use the study and the table before question 6.

Mark scheme

1 Describe with figures (3)

  • Overall pattern: as walking time increases, the relative risk of CVD falls (a negative correlation) (1)
  • A worked comparison: for example, the relative risk is 0.28 lower (28% lower) with 60 or more minutes per day than with under 10 minutes (1)
  • The fall gets smaller at the top end: 0.12 between each of the first three groups, but only 0.04 from 30 to 59 minutes to 60 or more (1)
  • DO NOT ACCEPT a wording that says walking causes the fall, for example "walking leads to less CVD": the data show a correlation
  • Relative risk has no unit
  • A value copied from the table with no comparison does not earn the second mark

2 Percentage difference (2)

  • (0.88 − 0.72) ÷ 0.88 × 100 (1)
  • 18.2 (%) (1)
  • A correct answer with no working gets both marks
  • 18.18 or 18 gets 1 mark only: the question asks for 1 decimal place
  • DO NOT ACCEPT 28%: that compares the group with the baseline of 1.00, not with the 10 to 29 minutes group

3 Is the headline supported? (3): max 3

  • Relative risks of 0.76 and 0.72 mean a risk 24% and 28% lower than the baseline: about a quarter lower (1)
  • Halving the risk would need a relative risk of 0.50 (1)
  • Even the lowest ends of the error bars (0.68 and 0.62) are above 0.50, so the data do not support "halves": the headline exaggerates (1)
  • The study shows a correlation, so it cannot show that walking is the cause of the lower risk (1)
  • DO NOT ACCEPT a link between the error bars and reliability
  • Max 3

4 No pattern (3)

  • No. There is no pattern: the relative risk stays close to 1.00 (0.98, 1.02 and 0.99 for the tea groups) (1)
  • All three error bars include 1.00 (0.90 to 1.07, 0.93 to 1.12 and 0.87 to 1.13) (1)
  • So no tea group differs significantly from the baseline: the data do not support the statement (1)
  • ACCEPT "the values go up and down" with a figure for the first mark
  • A statement that there is no pattern is as creditable as a statement of a pattern

5 Why not a cause? (4): max 4

  • The sample was volunteers who joined a fitness app, so they are likely to be health-conscious (1)
  • So the lower risk could come from other healthy habits, and the sample is not representative of the population (1)
  • Smoking and diet were not recorded or controlled (1)
  • People who walk more may also smoke less or eat better, so a confounding variable could explain the lower risk (1)
  • The study lasted only 3 years, and CVD develops slowly (1)
  • So few cases were counted, or an effect over a longer time cannot be seen (1)
  • People already in poor health may walk less (cause the other way round) (1)
  • So the CVD could be the reason for the low walking, not the result of it (1)
  • Each reason needs its explanation: the first line of each pair is the reason, the second the explanation
  • DO NOT ACCEPT a general "correlation is not causation" with no reason from this study
  • Max 4

6 Remove a confounding variable (2)

  • Compare night-shift and day workers who are all non-smokers (and all smokers) separately / match the groups for smoking (1)
  • If the difference in CVD stays, smoking does not explain it all; if it goes, smoking explains it (1)
  • ACCEPT a comparison of CVD rates within each smoking group, or a statistical adjustment for smoking
  • IGNORE "do a bigger study" with no change to the comparison

7 To what extent? (6): marked by level

Levels for question 7
LevelMarksWhat the answer looks like
35 to 6Data used with figures. Limitations explained, with what each one means for the statement, and a judgement that follows. 5 marks: what one limitation means for the statement. 6 marks: what two limitations mean.
23 to 4Data described, and limitations given, but what they mean for the statement is missing or thin. 3 marks: one limitation. 4 marks: two limitations.
11 to 2Descriptions of the data only. 1 mark: one description. 2 marks: two descriptions, for example the CVD rates and the smoking figures.
  • Indicative content, AO2 (from the data):
  • CVD cases are higher in night-shift workers: 27 against 18 per 1 000 people, which is 9 more, or 50% higher
  • The mean ages are similar (44 and 45 years), so age is unlikely to explain the difference
  • The groups are large (4 800 and 1 200) and the study lasted 10 years
  • Smokers are 38% of night workers but 21% of day workers: smoking is a confounding variable
  • Diet and exercise were not recorded, so they may differ between the groups
  • There are no error bars or statistical test, so we do not know whether the 9 per 1 000 is significant
  • Indicative content, AO1 (from knowledge):
  • The study shows a correlation; a cause needs more: controlled comparisons or repeated results in other populations, and a mechanism
  • Smoking is a risk factor for CVD, so the extra smokers could produce some or all of the extra cases
  • The unequal group sizes do not matter, as the cases are given per 1 000 people
  • NB A Level 3 answer uses figures, gives limitations, says what each limitation does to the statement, and ends with a judgement such as "partly supports"
  • ACCEPT other valid limitations
  • An answer that only describes the table stays in Level 1

Self-marking checklist 5 min

Before you hand in, check that:

  • every comment on data has the overall pattern first, then a comparison worked out from two figures;
  • you wrote "correlation" or "linked to", never "causes", unless the question asks for a cause;
  • every limitation says what it does to the conclusion, not just that it exists;
  • error bars are linked to significance, not reliability;
  • for "to what extent", the last sentence is a judgement.

After you hand in, the mark schemes open. Then:

  • tick each mark point you earned: a limitation with no link to the conclusion earns less;
  • for each missing figure, unit or link, add it in a different colour;
  • write your total out of 23, and one target, for example: "I will say what each limitation does to the conclusion."

Back to the lesson

Was this page useful?

Goes straight to the two people who make these pages. Please don't include students' names or other personal details.