Comment by valenaut | Hacker News Reader

Comment by valenaut | Hacker News Reader

Parent Full thread

Many epidemiologists are effectively specialized data scientists, and I don't think a random data scientist could provide equal analysis. Sure, they can both run a regression, but you benefit from understanding particular study designs, as well as biology/chemistry/toxicology.

I don't know much about infectious disease epidemiology, but I'll use glyphosate toxicity as an example from environmental epidemiology. Glyphosate was found to be non-carcinogenic by the EPA, but if you review the literature on glyphosate toxicity, industry-affiliated researchers tended to study pure glyphosate, and these studies were the majority of those considered by EPA. Researchers without industry affiliation tended to study glyphosate in the formulations actually sold in products such as RoundUp, and found much stronger evidence of carcinogenic toxicity.

I don't think a generic data scientist would think to ask the question about glyphosate formulations, as many people at EPA did not.

I agree, though my impression is epi's use policy driven hypothesis' to construct data driven stories around, which is great for directing their attention to things like your example, as ultimately it was an activist narrative backed by epi data that made people aware of how bad some of those carcinogens were.

However, there are many epis who are just political scientists, and calling them epidemiologists and conflating them with bio or medical experts is sort of like the way we talk about devs as engineers and architects. Covid has really put them in the spotlight, as they're the ones driving policy, which becomes a big deal when we start using their work as justification for internal passports and mandates. I'm saying the epis working on covid aren't held to the standards of other scientists or even data science hackers, and they are being used to launder radical policies through "expert" models data. I've got massively controversial opinions on these topics because of my health information privacy and security work, so YMMV.

I'd posit that a data scientist with working professional competence in say, R and Pandas has as much or more modelling experience for empirical discovery than most epi's where I'd suggest the epi's are a kind of policy profession, and that someone with those basic quantitative skills is equipped to question policies and assertions produced by said epis.

Yes, there are policy-focused epidemiologists and informatics-focused epidemiologists. My partner is the latter, so I'm much more familiar with that world, and that's what I associate with "epidemiology". There is overlap, of course, but I take your point that you can't necessarily know someone's quantitative analysis skills from knowing they are an epidemiologist.

I would just note that "a data scientist with working professional competence in say, R and Pandas" describes a good number of epidemiologists, who also have infectious disease, toxicology, and study design knowledge on top of this.

Reply on news.ycombinator.com