Partial Correlation Analysis

Partial correlation analysis measures the linear association between two variables after removing the linear effect of one or more control variables from both of them.

편상관분석은 하나 이상의 통제변수의 선형 영향을 두 변수 모두에서 제거한 뒤 두 변수 사이의 선형 연관성을 측정하는 방법입니다.

How do I measure the correlation between two variables while controlling for a third variable?

제3의 변수를 통제한 상태에서 두 변수의 상관은 어떻게 구하나요?

Partial correlation estimates how strongly two continuous variables are linearly related once the influence of one or more control variables has been removed from both. It fits when a zero-order correlation might be inflated or masked by a shared confounder, such as age or income.

When to use Partial Correlation Analysis

• Two continuous variables that may both be driven by a common covariate
• Checking whether a correlation persists after adjusting for confounders
• Building a matrix of conditional associations among several continuous variables

When to use something else

• Control variable should be removed from only one of the two variables: use semipartial correlation
• Ordinal or non-normal variables with monotonic relationships: use Spearman partial correlation

Assumptions and requirements

• Continuous variables
• Linear relationships among all variables
• Approximate multivariate normality for significance tests
• No extreme outliers

What the analysis reports

• Partial correlation coefficient
• Degrees of freedom and p-value
• Confidence interval
• Comparison with the zero-order correlation

How to report it

Controlling for [covariate], [variable 1] and [variable 2] were positively correlated, r([df]) = [value], p = [value].

Related methods

• Pearson Correlation
• Semipartial Correlation
• Multiple Linear Regression

Run Partial Correlation Analysis on Quorant