Least Squares Fit of Definite Quadratic Forms by Convex Programming

This paper considers the problem of fitting a quadratic regression law subject to the condition that the fitted surface be convex (concave). A computational algorithm is described for determining the constrained regression coefficients using an existing non-linear programming algorithm. Some of the properties of the estimators are described.

• 저널Management Science · 13(11) · 913–925
• 토픽Advanced Statistical Methods and Models · Statistics and Probability
• DOI10.1287/mnsc.13.11.913
• 저자H. O. Hartley, R. R. Hocking, William P. Cooke