Identifying the significant factors in treated water using binary logistic regression model : a class study in Perlis, Malaysia
Abstract
The production of disinfection by-product (DBP) is result of reaction between
disinfection agent with the natural organic matter, bromide or iodide in the water.
Trihalomethane (THM) is the main group of DBP found in many chlorinated water. THM
consist of trichloromethane (TCM), dibromochloromethane (DBCM),
dichlorobromomethane (DCBM) and tribromomethane (TBM). In this study, data of
treated water taken from year 2015 in state of Perlis, Malaysia. For sampling point, Perlis
was divided into four district (Kangar, Jejawi, Kuala Perlis and Arau) and water quality
variables measured are water pH, water temperature, residual chlorine, TCM readings,
the purified water and non purified water. A binary logistic regression was proposed to
predict the probability of purified water in housing area using all the continuous variables
measured. Correlation between TCM formation and these variables were estimated using
a Spearman’s correlation test. A negative moderate correlation exist between residual
chlorine and TCM in Arau (p= -0.537, p<0.05) . Besides that, there exist a strong
negative correlation between water pH and residual chlorine in Jejawi (p= -0.603, p<0.01)
but no correlation exist between pH and TCM. Same result
obtained for district Kangar (p= -0.722, p<0.01) . Additionally, when correlation
analysis did on the overall data, there exist a negative weak interaction between water
temperature and TCM formation in Perlis (p= -0.222, p<0.05) . However, some
variables such as pH and residual chlorine did not show any influence towards TCM
formation. Results of adjusted Welch’s F test suggesting that there is significant mean
different of TCM level in every district in Perlis. Significant pairwise difference obtained
between Kangar (x~= -0.189, SD=0.171) with Kuala Perlis (x~= -0.0478, SD=0.0667)
with mean difference of 0.14095. Besides that, Jejawi was (x~= -0.1852, SD=0.0816)
significantly higher of TCM mean compared to Kuala Perlis with mean difference of
0.1374 and Arau (x~= -0.092, SD=0.063) with mean difference of 0.09335. From binary
logistic regression model, residual chlorine and TCM have influence on the purified water
prediction. On the overall, the average TCM concentration in Perlis is 0.13mg/L (130
ppb) which is under acceptable maximum value of TCM while other THM such as
DBCM, DCBM, and TBM are non-quantifiable because of concentration below detection
limit.