Glad you read it, but very confidently disagree. ANOVA is exactly what you use looking for a similar effect in pops that could have different means but otherwise similar distributions, non-parametrics explicitly are about ordinal scales, and the power of the combined population is greater than either alone.