a type of regression analysis in which independent variables are used to predict the rank (as opposed to the actual value) of a dependent or response variable. This approach is particularly useful for distributions of errors that lack normality in their end values. See nonparametric regression.