The dry weight and specific stem weight of seedlings sampled
from the seedbeds were arcsine (square-root) transformed and
analyzed using a 2-way univariate General Liner Model (GLM). The
seedbed-blocks were initially included in the analyses but were
later removed because block had no effect. Data from the choice-
experiment that was conducted in the greenhouse were ranked
because of non-independence of observations. The ranked data
were subjected to repeated-measures GLM with age, snail size,
seedlings per hill, and seedbed seed-density as main factors and
removing block (which had no effect) from the analysis. The con-
trolswere removed fromthe analysis because survival of hills in the
control arenas was 100% in every case. Hill biomass and absolute
snail-related reductions in hill biomass (differences between hill
biomass in arenas with snails present and control arenas without
snails) were estimated for each block and analyzed using GLM.
Block had no significant effect and was subsequently removed from
the analysis. Data for 2e13 days of exposure were analyzed using
repeated-measures GLMas a randomized block design. Subplot and
plant characteristics (survival, sample weight, hill weight, number
of tillers per hill) at the end of the experiment (32 days) after
destructive sampling were analyzed using GLM (assuming that
snails had unrestricted or reduced (in the case of hand-picked
plots)) access to all plots. Survival data were arcsine (square-
root)-transformed. Analyses were conducted using SPSS (version
18). Residuals were plotted after each analysis and found to be
normal and homogeneous.