Estimating peak flow characteristics at ungaged sites by ridge regression

Open-File Report 82-333
By:

Links

Abstract

A regression simulation model, is combined with a multisite streamflow generator to simulate a regional regression of 50-year peak discharge against a set of basin characteristics. Monte Carlo experiments are used to compare the unbiased ordinary lease squares parameter estimator with Hoerl and Kennard's (1970a) ridge estimator in which the biasing parameter is that proposed by Hoerl, Kennard, and Baldwin (1975). The simulation results indicate a substantial improvement in parameter estimation using ridge regression when the correlation between basin characteristics is more than about 0.90. In addition, results indicate a strong potential for improving the mean square error of prediction of a peak-flow characteristic versus basin characteristics regression model when the basin characteristics are approximately colinear. The simulation covers a range of regression parameters, streamflow statistics, and basin characteristics commonly found in regional regression studies.
Publication type Report
Publication Subtype USGS Numbered Series
Title Estimating peak flow characteristics at ungaged sites by ridge regression
Series title Open-File Report
Series number 82-333
DOI 10.3133/ofr82333
Edition -
Year Published 1982
Language ENGLISH
Publisher U.S. Geological Survey,
Description [iii], 155 p., ill. ;28 cm.
Google Analytic Metrics Metrics page
Additional publication details