Intensity of infection (no. of metacercariae/fish) and density of snails
was analyzed using negative binomial regression [20] using various
pond characteristics as predictors and standard errors were adjusted
for clustering within ponds. Similarly, FZT infection status of individual
fish was analyzed using logistic regression. Both these analyses were
specified in generalized linear models with a log and a logit link function,
respectively and standard errors were adjusted for ponds as clusters.
Several predictors were tested individually and later significant
predictors were tested jointly using a stepwise selection procedure
(p for removal = 0.10 and p for entry = 0.05). Scatter plots between
prevalence of FZT infection, arithmetic mean intensity of infection in
fish, density of host snails and pond depth were made to illustrate the
associations. Linear regressionwasmade on the data and to test for possible
geographic clustering we also did the analysis with commune as a
random factor.