User:Mdougher/sandbox

General Monotone Model
In statistics, the General Monotone Model (GeMM) is a semi-parametric multiple regression procedure in which model parameters (regression coefficients) are estimated that maximize the rank-ordered relationship between the dependent variable and a set of predictors. GeMM finds the regression coefficients that maximize the value of Kendall's tau, a rank-order correlation statistic, between the fitted values and the data. This contrasts with standard least-squares regression techniques in which model parameters are estimated minimize the squared distance between the fitted values and the data. The approach used in GeMM is similar to the Maximum Rank Correlation (MRC) estimator.

Recent implementations of the General Monotone Model implement Order Constrained Least-squares Optimization (OCLO) in which the least-squares solution is obtained conditioned on the model that maximizes ordinal fit.