Statistically significant factors for each of the models were
identified through numerical and graphical techniques. The
factor selection procedure through effect analysis using JMP is
demonstrated in Table 3. Table 3 shows the statistical significance
ofmain factors and interaction terms onTHMsformation
model in the PP (model 2 in Table 4). THMs at the WDS, free
residual chlorine, temperature in the PP and interaction of TOC
and temperature were found to be statistically significant for
modeling THMs in the PP (Table 3). Despite TOC was not identified
to be a significant main factor, TOC was included in the
model because the interaction of TOC and temperature was
identified to be significant. Inclusion of TOC was necessary to
maintain statistical consistency (Montgomery and Runger,
2007). In this sampling program, reaction time was considered
as constant in all cases as thesamples fromtheWDSwere taken
in the late evening after the last use of water during night and
the samples fromthe tapwater in house (representing PP) were
collected in the early morning of next day prior to the first use.
As the reaction time was maintained to be constant, this factor
was not included in the model. The effects of parameters were
also tested using graphical techniques (Normal plots, interaction
plots, etc.: not shown) and the significant factors for
modeling were identified. Details about the graphical techniques
for the selection of significant factors can be found in
Chowdhury et al. (2010). Further details on the screening analysis
may be obtained from Montgomery and Runger (2007).