whether the number of variants n should be replaced by 1+r? Esp. considering the current formula will give a smaller sample size with an increasing split ratio r, ie. an unequal split would require a smaller sample size to reach the same error rate, which is not so intuitive?
Hi all,
I'm wondering in the sample size formula,
https://github.com/zalando/expan/blob/master/expan/core/statistics.py#L168
whether the number of variants
n
should be replaced by1+r
? Esp. considering the current formula will give a smaller sample size with an increasing split ratior
, ie. an unequal split would require a smaller sample size to reach the same error rate, which is not so intuitive?Thanks!