2.5. Data analysis
Residual plots were used to explore normality and homogeneity of
data. Percentage data were arcsine transformed prior to further statistical
analysis. Differences in mean biochemical composition between
densities and instars were compared using two-way analysis of variance
(ANOVA) and significant differences identified using Tukey's
HSD tests for post-hoc multiple comparisons. Where no interaction between
two factors occurred, differences between densities or instars
were reported. Instar 9, 12, and puerulus samples are presented on
graphs but were excluded from analyses because there was insufficient
replication (no LD replicates for instar 9; only one LD replicate for instar
12; and no HD replicates for pueruli, resulting in unbalanced analyses).
Survival and growth measurements were compared between densities
using an independent sample t-test. Exponential regression was used to
describe the relationship between growth parameters and instars. Linear
and power regression was used to analyse the relationship between
growth parameters and culture day using SigmaPlot version 12 (2011
Systat Software Inc.). ANOVA and t-tests were performed using SPSS
version 16.0 (2007 SPSS Inc.). The level of significance for all analyses
was determined at Pb0.05. Data are presented as mean±standard
error (SE) unless stated otherwise.