<p>
</p>
<p>Wouldn’t a better null hypothesis be that Berkeley is at the mean? I mean, we usually want to base our null hypothesis on the idea that at the outset of the experiment no difference exists between the two groups for the study? We’re trying to see if Berkeley is different from the national average, because that’s actually a testable hypothesis. It’s after that that we’ll go ahead and design a model that tells us the magnitude.</p>
<p>I argue that the data is all heteroskedastic anyway, poorly sampled, and pretty much useless.</p>