2.3. Model development
Three types of linear models (main factors; main factors,
interactions and higher orders; logarithmic) and two types of
nonlinear models (three parameters Logistic and four
parameters Weibull) were investigated to predict THMs and
HAAs in the PP and HWT. The main factor linear models are
the simplest form of multiple linear models (Montgomery and
Runger, 2007). These models are constructed using significant
main factors where interaction and higher order terms are
ignored. The parameters (e.g., model coefficients) and the
factors (predictor variables) are linear in this model. In the
main factor, interaction and higher order terms models, the
significant main factors, the effect of two factors varying
together (e.g., TOC and temperature) and higher order terms
(e.g., quadratic, cubic) are incorporated. However, model
parameters are linear. In the case of the logarithmic linear
model, values of the factors are transformed into a logarithm
and then linear regression is performed. The model parameters
and logarithm of factors are linear in these models. The
fitness and performance of the regression models are generally
estimated by the coefficient of determination (R2), F Ratio,
root mean square error (RMSE), significance probability, lack
of fit test, Normal probability plot of residuals and residuals
versus predict and data order plots (Montgomery and Runger,
2007).