Services on Demand
Journal
Article
Indicators
Related links
Similars in SciELO
Share
Agrociencia
On-line version ISSN 2521-9766Print version ISSN 1405-3195
Abstract
CAMPOS-ARANDA, Daniel F.. Transference of hydrologic information through multiple linear regression, with best predictor variables selection. Agrociencia [online]. 2011, vol.45, n.8, pp.863-880. ISSN 2521-9766.
It is necessary to have long records of annual hydrological data to get a truer picture of their variability, as well as reliable estimates of their statistical properties. To obtain these records it is common to use additional sources of data and transfer techniques. One technique is the multiple linear regression whose numerical application implies the optimum selection of close lengthy records (regressors) to have the extension of short registration be a reliable estimate. This selection process involves three analyses: 1) how to define the best estimates, 2) what regression equations should be investigated, and 3) which model has better predictive ability. For the first analysis four criteria based on the sums of the squares of the residuals are presented; for the second all possible regressions are investigated since in the problems of hydrological information transfer, we will have five regressors at the most; for the third, about selecting the best predictive model, we used the residual analysis and cross-validation. The numerical application described is an extension of the annual runoff volume record in the Platón Sánchez hydrometric station of the Tempoal river system in the 26 Hydrological Region (Pánuco, México). Here we used four regressors that are the records of other gauging stations in such system. We came to the conclusion that even in problems with multicollinearity, the selection criteria and analysis led to consistent results and allowed for the best regression equations. The similarity of the results obtained with the selected regression models generated confidence in the estimates adopted.
Keywords : residual mean square; multicollinearity; residual analysis; Durbin-Watson test; cross-validation; Rio Tempoal.