should lucas take the pill?
Today and tomorrow, I’ll respond to Sam Lucas’ argument on unrepresentative samples that was published in the journal Quantity and Quality. You can read it, for free, here. It’s good, so I recommend that you all take a look.
Today, I start with a not-so-crazy hypothetical question about samples and statistical inference. It goes like this:
Professor Lucas is taking a drug, “Berkeleyflaxin,” (Bf) for a very serious heart condition. His doctor said that Bf prevents heart attacks in 75% of patients who use the drug and have that specific condition. Later, Professor Lucas finds out that the clinical trials for Bf were based on a non-representative sample. Specifically, like many drug trials, no African-Americans were included in the trial, which is an issue given Professor Lucas’ personal background. The research was based on a convenience sample of all-White volunteers in the mountains of Utah. Let’s further stipulate, like many real drugs, that Bf is expensive and has a serious side effect, like nausea, vomiting, or increased risk of stroke. That means that taking Bf and hoping for a placebo effect is a costly option. You really have to choose between taking it and not taking it.
Should Professor Lucas take the pill? Show your work! Tomorrow, I’ll summarize Professor Lucas’ article on non-representative samples and give you my opinion of what he should do in this hypothetical.