If the road is not yet clear to you,

we will show you the way how to get through.

Partial correlation is by definition the unique correlation between two variables. In Venn diagrams it is the overlap between two circles without the overlap of the others.

Suppose three variables (all measured at an interval or ratio scale) are used in a research. Between these variables correlations can be computed. If there is no correlations between two variables (that is rx₁x₂ = 0), a Venn diagram will show no overlap. If there is a perfect correlation between two variables (that is rx₁x₂ = 1), a Venn diagram will show a complete overlap. A correlation between 0 and 1 will show a partial overlap.

Between 3 variables some overlap will be seen between all three variables. Some part will be unique. The unique part is the partial correlation coefficient. As a matter of fact three partial coefficients can be computed. The notation of the partial correlation is a small r with subscripts like r_12.3. Notice the dot between 12 and 3. This means the correlation is computed between variable x₁ and x₂ controlled for x₃. Now three different types of partial correlation can be computed: r_12.3, r_13.2 and r_23.1.

The usefulness of partial correlations becomes clear in regression analysis. In these analysis partial correlations are calculated for adding a predictive variable or a group of predictive variables.

Related topics to 'partial correlation'

• Variable 
  • Independent and dependent variables 
  • The level of measurement: nominal, ordinal, interval, ratio 
• Correlation
• Multiple correlation 
• Univariate regression 
• Multiple regression 
• Hierarchical regression 
• Mediation 
• Moderation

Analyse your data properly and quickly with our SPSS-tutorials.

Return to the index of this dictionary