Resumo

CAMPOS-ARANDA, Daniel F.. Regional transference of hydrologic information through multiple linear regression of ridge type. Agrociencia [online]. 2013, vol.47, n.5, pp.411-427. ISSN 2521-9766.

When annual long records are used of runoff, rainfall or flooding of a region with similar hydrological response, to amplify short series through the statistical technique of multiple linear regression (MLR), it is likely that those records by reason of their intrinsic similarity will lead to a problem of multicollinearity. This problem should be detected and quantified to know if it is acceptable, moderate, strong or serious and look for alternative solutions to the fitting method of the MLR by least squares of the residuals. In this study a diagnostic was made of multicollinearity through variance inflation factors and condition indices based on the eigenvalues. In addition, the biased fitting of the MLR is presented as an alternative method, known as Ridge regression. A numerical application in the system of the Tempoal river, of Hydrological Region No. 26 (Pánuco, México), was described to complete the short record of runoff volumes of the Platón Sánchez hydrometric station, based on the other four measuring stations that have long records. It is concluded that the principal advantages of Ridge regression are the ease of handling of transference with six or more regressions and the simplicity of its implementation and development by means of the Ridge trace.

Palavras-chave : multicollinearity; variance inflation factors; eigenvalues; eigenvectors; condition indices; Ridge trace; regional homogeneity; Tempoal River.

        · resumo em Espanhol     · texto em Espanhol     · Espanhol ( pdf )