Relationship without the influence of a third number (partial correlation)
In brief
Two numbers are related, but a third one may influence both. Example: shoe size and reading test in primary school children are related because both rise with age. The partial correlation asks: how strong is the relationship at the same age?
How it works
- "New analysis" → "Are two things related?".
- Choose the two number columns.
- Under "Should anything be controlled for?" choose one or more number columns. Without a choice it stays the simple correlation ("Continue without controlling").
What Statable calculates
- The linear influence of the chosen columns is removed from both numbers. Statable chooses Pearson or Spearman as for the simple correlation; there is no Kendall here.
- Only people with answers in all columns count. The page says how many that is and blocks if there are too few or if a controlled column is itself part of the question.
- PSPP cannot do this; Statable calculates it itself.
Reading the result
Next to the partial correlation (symbol pr, for Spearman prs) the simple correlation of the same people is shown. This shows what controlling changes: if the relationship shrinks a lot, it depended mostly on the third number. If it barely changes, that number plays no big role. Only what was measured and chosen is removed; other influences remain.
pr(117) = .18, p = .052, 95% CI [−.00, .35], controlling for age