Correlation

From Wikipedia, the free encyclopedia

Jump to: navigation, search

This article includes a list of references, but its sources remain unclear because it has insufficient inline citations. Please help to improve this article by introducing more precise citations where appropriate. (November 2010)

In statistics, multiple correlation is a linear relationship among more than two variables. It is measured by the coefficient of multiple determination, denoted as R², which is a measure of the fit of a linear regression. A regression's R² falls somewhere between zero and one (assuming a constant term has been included in the regression); a higher value indicates a stronger relationship among the variables, with a value of one indicating that all data points fall exactly on a line in multidimensional space and a value of zero indicating no relationship at all between the independent variables collectively and the dependent variable.
Unlike the coefficient of determination in a regression involving just two variables, the coefficient of multiple determination is not computationally commutative: a regression of y on x and z will in general have a different R² than will a regression of z on x and y. For example, suppose that in a particular sample the variable z is uncorrelated with both x and y, while x and y are linearly related to each other. Then a regression of z on y and x will yield an R² of zero, while a regression of y on x and z will yield a positive R².

[edit] Fundamental equation of multiple regression analysis

The coefficient of multiple determination R² (a scalar), can be computed using the vector c of cross-correlations between the predictor variables and the criterion variable, its transpose c', and the matrix R_xx of inter-correlations between predictor variables. The "fundamental equation of multiple regression analysis"^[1] is

R² = c' R_xx⁻¹ c.

The expression on the left side denotes the coefficient of multiple determination. The terms on the right side are the transposed vector c ' of cross-correlations, the inverse of the matrix R_xx of inter-correlations, and the vector c of cross-correlations. Note that if all the predictor variables are uncorrelated, the matrix R_xx is the identity matrix and R² simply equals c' c, the sum of the squared cross-correlations. Otherwise, the inverted matrix of the inter-correlations removes the redundant variance that results from the inter-correlations of the predictor variables.