Reviewer asserts “the appropriate test for these particular data is a two-way ANOVA, (but be careful when analyzing proportions. The data should be arc-sine transformed to minimize deviations from the assumption of normality).
ANOVA has long been used to classify treating a system in two different experimental conditions (like different temperatures). There are 3 key assumptions:
- the cases are independent (I believe this is the same sort of independence when we assert that different embryos are independent random draws from the underlying population distribution).
- the data are normal. (Our experimental data is not described by a normal distribution. A simple, biophysical model predicts that the data should not be normally distributed).
- homoscadsticity: the data has constant variance. (this is certainly not the case, and it is in fact one of the conclusions of our paper).
- required for unit-treatment additivity: observed response = unit response + treatment effect. There is no a priori reason to believe the effect of temperature on failure rate should be additive.