You are here:
CONFIDENCE INTERVALS FOR A CROP YIELD LOSS FUNCTION IN NONLINEAR REGRESSION
Citation:
Lee, E., D. Tingey, AND W. Hogsett. CONFIDENCE INTERVALS FOR A CROP YIELD LOSS FUNCTION IN NONLINEAR REGRESSION. U.S. Environmental Protection Agency, Washington, D.C., EPA/600/J-90/294 (NTIS PB91146506), 1990.
Description:
Quantifying the relationship between chronic pollutant exposure and the ensuing biological response requires consideration of nonlinear functions that are flexible enough to generate a wide range of response curves. he linear approximation (i.e., Wald's) interval estimates for ozone-induced relative crop yield loss are sensitive to parameter curvature effects in nonlinear regression. he adequacy of Wald's confidence interval for proportional response is studied using the nonlinearity measures proposed by Bates and Watts (1980), Cook and Goldberg (1986), and Clarke (1987 a & b) and the profile t plots of Bates and Watts (1988). umerical examples comparing Wald's, likelihood ratio, the bootstrap, and Clarke's adjusted 95% confidence intervals for relative crop yield loss are presented for a number of ozone exposure studies conducted by the National Crop Loss Assessment Network (NCLAN) program. t ambient levels of ozone concentration, the effects of nonlinearity were significant and invalidated the adequacy of Wald's confidence interval. epending upon the severity of the curvature effects, an alternative interval (i.e., Clarke's adjustment to Wald's interval or the likelihood ratio interval) for proportional yield loss should be considered.
