Unit 4 · Relationships
How to run a partial correlation and Kendall's tau in SPSS
Control for a third variable with partial correlation in SPSS, and use Kendall's tau-b for small samples with tied ranks.
Quick answer: How do you run a partial correlation and Kendall's tau in SPSS?
- Click Analyze › Correlate › Partial.
- Move the two variables into Variables and the control variable into Controlling for.
- Click OK and read the correlation and its df.
Report it in APA style: Controlling for motivation, revision hours were still related to exam score, r(47) = .31, p = .029.
Partial correlation and Kendall’s tau
A partial correlation measures the relationship between two variables after removing the influence of one or more others. It answers “is there still a link once we account for X?”. Kendall’s tau-b is another rank-based correlation, an alternative to Spearman.
When to use it
- Partial: you suspect a third variable (a confound) explains some of a relationship, such as motivation affecting both revision and grades.
- Kendall’s tau-b: small samples, or ordinal data with many tied ranks.
How it works
A partial correlation removes the part of each variable that the control variable predicts, then correlates what’s left. If the correlation shrinks a lot, the control variable explained much of the original link. Degrees of freedom are N − 2 − the number of control variables. Kendall’s tau compares every pair of cases and counts whether they’re ordered the same way on both variables; its values are usually smaller than Spearman’s for the same data.
Assumptions
- Partial: the same assumptions as Pearson, for all variables involved.
- Kendall: ordinal or Scale data, monotonic relationship.
Worked example
Revision and exam score correlated r = .55, but controlling for motivation the partial correlation fell to r(47) = .31, p = .029. Motivation explained part of the link, but revision still mattered.
Key terms
- Confounding variable
- A third variable related to both variables of interest.
- Zero-order correlation
- The ordinary correlation, with nothing controlled.
- Tied ranks
- Cases with the same value, which share a rank.
Common mistakes
- Treating partial correlation as proof that the remaining link is causal.
- Controlling for a variable caused by one of the others (a mediator), which can remove a real effect.
How to do it in SPSS
The SPSS syntax
Prefer typing to clicking? Open File › New › Syntax, paste this in, select it and choose Run › Selection. Every dialog's Paste button writes the same kind of code.
* Partial correlation (exam_hours.sav).
PARTIAL CORR
/VARIABLES=hours score BY motivation
/SIGNIFICANCE=TWOTAIL
/MISSING=LISTWISE.
* Kendall's tau-b (tutoring.sav).
NONPAR CORR
/VARIABLES=satisfaction sessions
/PRINT=KENDALL TWOTAIL NOSIG.Practice data
Download the file used in this lesson and run it yourself in SPSS: you should get the same results as the screens above.
- exam_hours.savRevision hours, exam score and motivation (N = 50)
- tutoring.savSatisfaction ratings and tutoring sessions attended (N = 40)
Now practise it interactively →
Reading is only half of it. Tap through the real SPSS menus, read the output and pick the right write-up in 6 quick questions, with instant feedback on every answer.
Practice for this lesson is part of Pro, with every unit, unlimited hearts and saved progress.