2.5. Statistical analyses
IBMSPSS, Version 20 for Windows was used for all statistical analyses.Homogeneity of variances and normality among mean values were assessed using Levene's test for equality of variance errors and the standardised residuals against the predicted mean plot, respectively. All percentage data was arcsine transformed before analyses. All variables were analysed using two factor ANOVA, with water temperature as the first factor and dietary
protein level as the second factor. When significant interactions were observed, post-hoc tests were used to detect significant differences between all treatment combinations (Student Newman–Keuls). Linear and second order polynomial regression analyses were also applied to SGR,feedconsumptionrate(mg individual−1day−1)andFCR.Asignificance level of P b 0.05 was used for all statistical tests. All values are presented as means ± standard error (SE) of the mean unless otherwisestated.