Second, in order to examine the relationship between the
occurrence of viruses or the abundance of bacteria and the
physicochemical environmental variables recorded, loglinear
(for bacteria abundance, with a quasiPoisson distribution)
or logit (for viruses occurrence, with a Bernoulli
distribution) models were separately fitted to each
response variable (abundance of TC, FC, and FE; occurrence
of rotavirus, enterovirus and astrovirus) with the
environmental variables (temperature, conductivity, pH
and Oxygen concentration) as predictors; for viruses'
models, the abundance of indicator bacteria (i.e., TC, FC
and FE) were also included as predictors. In each case, predictor
variables significantly related to the response variable
were identified also with basis on the change in
deviance that its deletion from the model produced.
Finally, simple two-way analyses of variance were used to
test for differences between years, between seasons and
between seasons within each year in the means of the
physicochemical environmental variables recorded (pH,
Oxygen concentration, temperature and conductivity).