where d is the average distance separating all pairs of closest points, and A is the area of a region covered by the n points.Values of RI close to 0 indicate that the points in question are clustered; values close to 1 indicate a geographically random (Poisson distribution) arrangement of points.
To also account for underlying assumptions of normality in the spatial data, \citet{Griffith2005} proposed a method in which the expected value of the estimated variance of the response variable is calculated in a regression model and using the variance inflation factor, and an effective sample size approximation is obtained as a function of the covariance structure of the model.