Comments on: E-Cig Study: Much Smoke, Little Light

By: St. Rev

St. Rev — Sun, 30 Mar 2014 08:38:11 +0000

The study is comparing e-cig users (a very small and rapidly changing demographic) to non-e-cig users (a large, relatively stable population). There are going to be many uncontrolled variables for which the two populations differ. Some but not all of these differences will be in factors that are rapidly changing within the overall e-cig demographic.

For example, a sample of e-cig users taken in 2011 will probably differ significantly in novelty-seeking[1], age, and motivation to quit from a sample of non-e-cig users taken in 2011. It will also probably differ in those measures from a sample of e-cig users taken in 2021–the population is likely to regress to the mean in a lot of ways by then.

1: There’s some reason to think novelty-seeking itself is correlated to propensity for addiction. That alone would render the whole study fishy.

Report comment

By: RCF

RCF — Sun, 30 Mar 2014 06:30:34 +0000

@Anonymous

Well, okay. But the “general population” isn’t really relevant, is it? Whether e-cigarettes are effective is properly understood with regard to their effect on people motivated enough to try a smoking-cessation aid, not their effects on the general population.

Once a person even thinks to ask “Will e-cigs help me stop smoking?”, we have a reasonable basis for believing that “people who want to stop” is a more representative group than “general population”.

Report comment

By: Damien

Damien — Fri, 28 Mar 2014 16:11:28 +0000

It’s deliberate that the hypothetical researcher said she wanted to study the rates among first-time smokers, then totally failed to screen her cig shop subjects for being first-time smokers?

Report comment

By: Damien

Damien — Fri, 28 Mar 2014 16:09:59 +0000

One party of young Lunars, if the Silver Pact helps bring them together, sure. A whole subculture of young Lunars is harder…

(In a game with two Abyssals, an Infernal, a Dragon-Blood, and a brand new Solaroidish Exaltation.)

Report comment

By: Damien

Damien — Fri, 28 Mar 2014 16:07:05 +0000

Does this apply if the follow-up study is simply looking at the people originally questioned, as in the hypothetical study? Seems to me it only applies if you’re comparing two population samples at different times, not the same sample over time.

Report comment

By: St. Rev

St. Rev — Fri, 28 Mar 2014 06:25:04 +0000

I think the general heading for this problem is “selection bias”.

The specific error in the Hypothetical World example is a mismatch between the researcher’s explicit intention to study first-time-smokers, and the methodology of sampling from the customers of a smoke shop. In the Real World study, it’s unlikely that e-cig users who declare an intention to quit are correctly matched to the wider smoking population.

Report comment

By: Scott Alexander

Scott Alexander — Fri, 28 Mar 2014 00:47:41 +0000

What Itai said. I do appreciate the comment and analogy.

Report comment

By: Itai Bar-Natan

Itai Bar-Natan — Thu, 27 Mar 2014 21:14:08 +0000

I think the remark is sarcastic but not critical. Scott Alexander is trying to make a light-hearted joke without any ill intent.

Report comment

By: Erik

Erik — Thu, 27 Mar 2014 18:49:10 +0000

I’m not sure how to take this remark. It looks vaguely like critical sarcasm, but what are you trying to be sarcastic about?

I shared what looks to me to be a similar case where intuitive reasoning about frequencies relative to observations and instances easily goes astray because of the uneven distribution of durations, and that’s a horribly clunky way of trying to describe what I see as a general factor, which is part of why I posted the anecdote: trying to see a less clunky way of describing the common factor, if there is one. I don’t for one moment think that the distribution of extraterrestrial wizards is the non-ivory-tower problem here.

Report comment

By: Anon

Anon — Thu, 27 Mar 2014 12:15:30 +0000

I don’t have access to the full article – does anyone know whether/how they measured cigarette use, or how they determined whether a user had “quit”?

I used to be a pack-a-day smoker, and now I use e-cigs. I’ve purchased two packs of cigarettes in the past three months and haven’t finished them. I’ve only smoked when I ran out of e-cig cartridges (due to shipping delays, mostly) or wanted to make smoking fetish clips (yeah, that’s a thing). Have I “quit” smoking? Depends on your criteria.

I’ve also noticed that when I *do* smoke (without also using e-cigs, e.g. because I’ve run out), I smoke fewer cigarettes: 5-7 per day, as opposed to the ~20 per day I smoked when I wasn’t using e-cigs. They seem more potent and less pleasant to use than they did before – I’m no longer used to the smell, for example, which I don’t really like.

Report comment