> but this test has a Gaussian assumption, not a Poisson assumption. In practice the Gaussian assumption might not be too badly abused (at least for Twitter) s

> but this test has a Gaussian assumption, not a Poisson assumption. In practice the Gaussian assumption might not be too badly abused (at least for Twitter) since the Poisson distribution approaches the Gaussian when lambda is large.