Partial Correlation Calculator

Partial correlation measures the linear relationship between two variables after removing the linear influence of another variable from both. An ordinary Pearson coefficient between X and Y can be large because both variables are associated with Z. A first-order partial correlation asks what relationship remains between X and Y when that shared association with Z is held constant. This makes it useful for preliminary confounding analysis, exploratory research, education, and checking statistical software output. The calculator requires three pairwise Pearson coefficients: the correlation between X and Y, the correlation between X and the control variable Z, and the correlation between Y and Z. It subtracts the portion of the X-Y association implied by their separate relationships with Z, then standardizes the remainder. The output stays between negative one and positive one when the three inputs form a valid correlation structure. A positive result indicates that larger residual values of X tend to accompany larger residual values of Y. A negative result indicates an inverse residual relationship. Magnitude describes strength, not importance or causation. A coefficient close to zero means little linear association remains after controlling for Z, while a value closer to either endpoint indicates a stronger remaining linear pattern. Context, sample size, measurement quality, and uncertainty still matter. This calculator reports the coefficient only; it does not calculate a confidence interval or significance test because those procedures also require the sample size and assumptions about the observations. The formula becomes undefined when either X or Y is perfectly correlated with Z. In that case, removing Z leaves no residual variation for that variable. Inputs should also be correlations computed from the same observations and using the same treatment of missing values. Pairwise coefficients assembled from unrelated samples may be mathematically inconsistent. Use the result as a concise descriptive adjustment, then inspect data plots and a suitable regression model when decisions depend on nonlinear effects, interactions, additional control variables, or causal interpretation.