IB Psychology HLTopic 5 — Data AnalysisPaper 3 & IAHL only~10 min read
What the Correlation Coefficient Tells You
You already know a correlation coefficient runs from −1 to +1. At HL you need more than that: which test produced it, how much of the variation it actually accounts for, and whether it beat the critical value. A coefficient on its own is a number without a verdict.
📚 What you need to know
The coefficient runs from −1 to +1. The sign gives the direction; the size gives the strength.
Spearman’s rho is used for ordinal data; Pearson’s r is used for interval data that meet the parametric assumptions.
Roughly: below 0.3 is weak, 0.3 to 0.7 is moderate, above 0.7 is strong.
r², the coefficient of determination, gives the proportion of variation the two co-variables share.
Significance depends on both the coefficient and the sample size — a small sample needs a much larger r.
The observed value must be compared with a critical value before it can be called significant.
No coefficient, however large, establishes cause.
What the size of r looks like
Numbers are hard to picture, so here are three real scatterplots at increasing strengths. Notice how much structure has to be present before the relationship is obvious to the naked eye.
The left-hand plot is worth staring at. Plenty of published psychological findings sit around this strength, which is why “significant” and “useful” are not the same word.
Spearman or Pearson?
Both produce a coefficient on the same −1 to +1 scale, and the choice between them follows the same logic as any other test decision.
Spearman’s rho
Pearson’s r
Data level
Ordinal, or interval data that fail the parametric assumptions
Interval or ratio
Distribution
No assumption of normality
Must be normally distributed
What it works on
The ranks of the scores
The actual values of the scores
Power
Less powerful
More powerful, so more likely to detect a real relationship
Reaction times, physical measurements, test scores in equal units
Because Spearman works on ranks rather than raw values, it will pick up any relationship where one variable rises as the other rises, even if the rise is not a straight line. Pearson is looking specifically for a straight-line relationship. That difference occasionally matters.
Squaring the coefficient
Here is the calculation that changes how you read every correlation you meet. Square the coefficient and you get r², the proportion of variation shared by the two co-variables.
Coefficient of determination
r = 0.6 → r² = 0.36, so the co-variables share 36% of their variation
That means 64% of the variation is down to something else entirely. A correlation of 0.6 sounds impressive until you realise nearly two-thirds of what is going on remains unexplained. Even a strong-looking 0.7 leaves 51% unaccounted for. This is the single most useful evaluation tool on this page, because it converts a vague-sounding number into a concrete claim about how much is left over.
Try it on the three plots above. r = 0.29 gives r² = 0.08, so 8% shared. r = 0.59 gives 35%. r = 0.94 gives 88%. Only the last of those describes a relationship strong enough to predict one person’s score from the other variable.
Strength is not significance
A coefficient of 0.5 might be significant in one study and not in another, because significance depends on the sample size as well as the strength. With a very small sample, even a fairly strong-looking correlation can easily arise by chance, so the critical value it must beat is high. With a large sample, a much weaker correlation clears the bar.
So the procedure is the same as for any inferential test. Take your observed coefficient, find the critical value using the number of participants, whether the hypothesis is directional, and your significance level. If the observed value equals or exceeds the critical value, the null hypothesis is rejected.
🧩 Reporting a correlation properly
Name the test and justify it from the data level.
Give the coefficient with its sign, and describe the direction in words.
Describe the strength using weak, moderate or strong.
Calculate r² and state the percentage of shared variation.
Compare with the critical value and state whether the result is significant, and at what level.
Add the cause warning. A correlation cannot establish that one variable causes the other.
Worked examples
WORKED EXAMPLE
Interpret a coefficient fully
A study of 40 students finds a correlation of −0.62 between hours of sleep and self-reported anxiety, measured on a 1 to 10 scale. Interpret this result.
Step 1: Name the appropriate test
Anxiety is rated on a scale, so the data are ordinal and Spearman’s rho is appropriate.
Step 2: Read the sign
The coefficient is negative, so as sleep hours increase, reported anxiety decreases.
Step 3: Read the strength
0.62 falls between 0.3 and 0.7, so this is a moderate relationship.
Step 4: Calculate the shared variation0.62² = 0.384, so the two co-variables share about 38% of their variation. Around 62% is explained by other factors.
Step 5: Add the limitation
No causal claim is possible. Anxiety could reduce sleep, sleep could reduce anxiety, or a third variable such as workload could drive both.
Moderate negative correlation; 38% shared variation; no cause establishedthe r² step is what lifts this from a description to an interpretation
WORKED EXAMPLE
Same coefficient, different verdict
Study A finds r = 0.45 with 12 participants and reports it as non-significant. Study B finds r = 0.45 with 90 participants and reports it as significant at p < 0.05. Explain how both can be correct.
Step 1: Note what is identical
The strength and direction are the same in both studies, and so is r² at about 20%.
Step 2: Explain what differs
Significance depends on the critical value, which changes with sample size.
Step 3: Give the mechanism
With only 12 participants, a coefficient of 0.45 could easily arise by chance, so the critical value is high and 0.45 does not reach it.
Step 4: Complete the comparison
With 90 participants, chance is far less likely to produce that pattern, so the critical value is much lower and 0.45 clears it.
Same strength, different sample sizes, different critical valuesthis is the clearest demonstration that strength and significance are separate ideas
💡 Exam tip
Always report direction and strength separately. The sign and the size answer different questions.
Use r² whenever you are asked to evaluate a correlation. It is the strongest single point available.
Justify Spearman or Pearson from the data level, exactly as you would any other test.
Say significance depends on sample size. It shows you understand what a critical value is for.
Finish every correlation answer with the cause warning, but explain it rather than reciting it.
Never call a correlation “high” without a number. Use weak, moderate or strong with the value.
⚠ Common mix-up
Reading a negative coefficient as a weak one. −0.9 is a very strong relationship.
Treating r as a percentage. It is r² that converts to a percentage of shared variation.
Assuming a significant correlation is a strong one. A large sample makes weak correlations significant.
Using Pearson’s r on rating scale data. Rating scales are ordinal, so Spearman’s rho applies.
Concluding cause from a strong coefficient. Strength changes nothing about direction of causation.
Forgetting the critical value comparison. An observed coefficient is not a verdict on its own.
Up next: Thematic Analysis of Qualitative Data — the final page of Topic 5, and a complete change of gear from numbers back to words.
Want this explained one-to-one?
Book a free session with an experienced IB Psychology tutor and get your trickiest topics made simple.