If we want to use the underlying ranks in our data and not be limited by the anova assumption (anova-assumptions), we can use the Mann-Whitney test.

1library(coin)
2wilcox_test(Time ~ IDE, data=ide2, distribution="exact")
3
4#

N.B.: Taking the log of time in the example of designing-for-experimental-control won’t change the result of a non-parametric test, because taking the log of the response doesn’t change the inherent ranks in the data.

References

• data-transformation-when-normality-is-violated
• variable-types
• data-distributions
• independent-samples-t-test
• confounding-variables
• one-way-anova