The HL Paper 3 skill: read an unseen study, interpret its data, and evaluate how far its conclusion holds. Applies across all four contexts.
4.4.1 Reading, Analysing, and Evaluating Data
What Paper 3 Tests
Paper 3 is HL, source-based: an unseen study you must read, analyse, and evaluate.
It draws on any of the four contexts, so it tests skill, not recall of named studies.
Two core moves: read the data accurately, then evaluate the evidence.
Reading Graphs
Bar chart = means across groups; line graph = change over time; scatterplot = relationship between two variables; box plot = spread (median + range).
Error bars show variability, drawn as SD or a 95% confidence interval.
Heavily overlapping error bars warn the difference may not be reliable; barely overlapping = more likely real.
Truncated axis (not starting at zero) exaggerates a gap. Always check where the axis begins.
Descriptive Statistics
Central tendency: mean (symmetric numerical), median (skewed / outliers), mode (categories).
Dispersion: range (crude); standard deviation = average distance of scores from the mean. Small SD = consistent group.
Outliers drag the mean but not the median, so skewed data are best summarised by the median.
Correlation Is Not Causation
Correlation coefficient r runs minus 1 to plus 1: positive, negative, or near zero. Further from zero = stronger.
Correlation is not cause. Three reasons two variables move together: one causes the other, reverse causation, or a third variable drives both.
Only a controlled experiment establishes cause. One of the most rewarded Paper 3 points on correlational data.
Example
Ice-cream sales and drownings correlate, but neither causes the other. Temperature drives both.
Significance vs Effect Size
P-value = probability of results this extreme if there were no effect. p < 0.05 = statistically significant, reject the null. Does NOT mean large or important.
A large sample size makes even a trivial difference significant.
Effect size = magnitude. Report both: is it real (significance) and big enough to matter (effect size)?
Qualitative Data and Credibility
Qualitative data = words; analysed by thematic analysis: codes grouped into themes.
Aim is depth, not statistical generalisation, so the test is transferability. Quality is judged by trustworthiness.
Techniques: triangulation, member checking, reflexivity, thick description, negative case analysis.
Answering Paper 3
Sequence: read the source (method, data type, sample) then interpret (state the finding), then evaluate, then reach a judgement.
To what extent asks for a judgement, not a list: develop two or three relevant issues and weigh them against strengths.