The forward stepwise mode used only six parameters (K, Fe, HCO3, NO3, pH and F) giving 74% correct assignation and the backward stepwise mode gave 72% correct assignation of cases using only one parameter (pH).
The forward stepwise mode used seven parameters (Mg, K, HCO3, Cl, NO3, SO4 and F) giving 90% correct assignation and the backward stepwise mode gave 89% correct assignation of cases using only three parameters (Mg, Cl and NO3).
Spatial backward stepwise mode Discriminant function analysis (DFA) discriminated five parameters viz., temperature, EC, NO3 N, Fe and Pb that cause variations among three groups with classification accuracies of 71.88% (Fig. 4).
Seasonal backward stepwise mode DFA identified sixteen important variables viz., pH, TDS, DOS, DO, ALK, PO4 3−, NO3 N, Cl, Ni, Cu, Mn, Co, Fe, Pb, Na and Mg as (Fig. 3; Table 2) with classification accuracies of 100% which bring variations in pre- and post-monsoon seasons.
A backward stepwise mode was used in order to avoid dropping non-significant variables that affected the model fitness.
The backward stepwise mode was used to avoid exclusion of non-significant variables that affected the model fitness [ 25].
Discriminant functions (DFs) and classification matrices (CMs) were derived from the standard, forward stepwise and backward stepwise modes of DA.
The result from the backward stepwise model was adopted.
We performed linear discriminant analysis (LDA) using three iterative approaches, two backward stepwise (a deterministic mode and a stochastic mode) and one forward stepwise.
In the present study, DA was carried out on raw data using three different modes: standard, forward stepwise and backward stepwise to construct discriminant functions (DFs) and to assess both temporal and spatial variations in groundwater quality.
Bacterial inactivation, as logarithmic reductions of Enterococcus spp. counts, was modeled through the backward stepwise procedure.
