The use of linear regression to obtain the parameter estimates for a nonlinear model could lead to imprecise results. This comment is to point out the problem of linearization in the parameter estimation for a nonlinear model. In Oke (2008), the reliability of ultimate BOD (L0) and BOD removal rate (k) were first calculated using the Weibull plotting position formula and then the L0-reliability and k-reliability data were fit to a nonlinear system-reliability model. Model parameters were estimated based on a linearized model and the linear regression. In this comment, we examine and compare the results of parameter estimation from the linear and nonlinear regressions. According to the residual plots, the nonlinear regression gives better estimates of the parameters than the linear one.
- Probability plot; linearization; nonlinear regression; residual plot