Data were transformed to improve normality using Box Cox transformation.
All total seed mass per square meter data were square root transformed to improve homoscedasticity, and error bars in the figures were back-transformed.
For SIS, data were log transformed to improve the normality of RIL distribution prior to QTL mapping.
For the identification of tolerant-specific metabolites, the normalised data was log10 transformed to improve normality and analysed by a two factorial ANOVA with interaction, where the factors were treatment (WW or D) and genotype.
The proportion of winged aphids was arcsine square-root transformed to improve normality and homoscedasticity.
Proportion data were arcsin-square root transformed to improve symmetry [49].
All variables were square-root or log10 transformed to improve normality and homogeneity of variance.
Data were natural-log transformed to improve the normality of the distribution.
Where necessary responses were appropriately transformed to improve the normality and constant variance regression assumptions.
Skewed variables were logarithmically transformed to improve normality before analyses.
Glucosinolate concentrations were log transformed to improve normality.
