Chi-Square and the Perils of NHST

The problem with $\chi^2$ is that what we usually care about when comparing distributions is the effect size, not the significance
We care about how unfair the die is, not whether it's perfect to 50 decimal places.
With enough data, $\chi^2$ will basically always be significant, no matter how trivial the difference

The problem with $\chi^2$ is that what we usually care about when comparing distributions is the effect size, not the significance
We care about how unfair the die is, not whether it's perfect to 50 decimal places.
With enough data, $\chi^2$ will basically always be significant, no matter how trivial the difference