Smoke Gets in Your IQs

There’s not much more revolting than the tobacco industry. Well, OK, there’s racist scum like Trump and Turnbull. And there’s greasy media apologists. And Bill Gates. And Mia Farrow.*

Alright, the world is full of awful people. But you get the point: it is difficult to be on the side of smoking and tobacco-pushing sociopaths.

Difficult, but not impossible.

Recently, the media was full of shock and horror at a new study on smoking. It was widely reported that 2/3 of people who try one cigarette end up as “daily smokers”. This was the conclusion of a meta-analysis, covering over 200,000 respondents from eight surveys. Professor Peter Hajek, one of the study’s authors, noted the meta-analysis constituted documentation of the “remarkable hold that cigarettes can establish after a single experience.”

Which is crap, and obvious crap. The implied suggestion that a single cigarette can turn a person into a helpless addict is nothing but Reefer Madness madness.

How can a respected and sophisticated academic study come to such a conclusion? Well, it doesn’t.

Anyone who has read the great debunking by Susan Traynor‘s son knows to never take a statistical study, much less a one sentence summary of a study, at face value. In this case, and as the authors of the study properly and cautiously note, that “2/3 of people” hides a wide variance in survey quality, response rates and response types.

More fundamentally, and astonishingly, the study (paywalled) never attempts to clarify, much less define, the term “daily smoker”. How many days does that require? The appendix to the study suggests that only three of the eight surveys included in the meta-analysis asked about “daily” smoking with specific reference to a minimal time period, the periods being 30 days, “nearly every day” for two months, and six months.

Of these three studies, the 2013 US NSDUH survey, which used the 30-day period, had around 55,000 respondents and the highest response rate, of around 72%. Amongst those respondents, about 50% of those who had ever smoked had at some time been “daily smokers” (i.e. for 30 days). Hardly insignificant, nor an insignificant time period, but a significant step down from “2/3 daily smokers”. (For some reason, the figures quoted in the meta-analysis, though close, are not identical to the figures in the NSDUH survey; specifically the number of people answering “YES” to the questions “CIGEVER” and “CIGDLYMO” differ.)

Even accepting the meta-analysis as sufficiently accurate, so what? What does it actually indicate? Reasonably enough, the authors suggest that their study has implications for efforts to stop people becoming regular smokers. The authors are tentative, however, rightly leaving the policy analysis for another forum. In particular, in the study the authors never make any claim of the “remarkable hold” that a single cigarette can have, nor do they make any remotely similar claim.

The “remarkable hold” line, which was repeated verbatim in almost every news report, originates from a media release from Hajek’s university. Of course barely any media organisations bothered to look beyond the media release, or to think for half a second before copying and pasting.

There is indeed a remarkable hold here. It is the remarkable hold university media units have on news organisations, which don’t have the time or experience or basic nous to be properly skeptical of the over-egged omelettes routinely handed to them on a platter.

Update: Just a quick addition, for those might doubt that Turnbull is racist scum.

* Yeah, yeah, no one knows, except Mia and Woody. But I believe Moses.

A Madness for all Seasons

Our fourth post on the  2017 VCE exam madness will be similar to our previous post: a quick whack of a straight-out error. This error was flagged by a teacher friend, David. (No, not that David.)

The 11th multiple choice question on the first Further Mathematics Exam reads as follows:

Which one of the following statistics can never be negative? 

A. the maximum value in a data set

B. the value of a Pearson correlation coefficient

C. the value of a moving mean in a smoothed time series

D. the value of a seasonal index

E. the value of a slope of a least squares line fitted to a scatterplot

Before we get started, a quick word on the question’s repeated use of the redundant “the value of”.


Now, on with answering the question.

It is pretty obvious that the statistics in A, B, C and E can all be negative, so presumably the intended answer is D. However, D is also wrong: a seasonal index can also be negative. Unfortunately the explanation of “seasonal index” in the standard textbook is lost in a jungle of non-explanation, so to illustrate we’ll work through a very simple example.

Suppose a company’s profits and losses over the four quarters of a year are as follows:

    \[ \begin{tabular} {| c | c | c | c |}\hline {\bf\phantom{S}Summer \phantom{I}} &{\bf\phantom{S}Autumn \phantom{I}} &{\bf\phantom{S}Winter \phantom{I}} &{\bf\phantom{S}Spring \phantom{I}} \\  \hline {\bf \$6000} & {\bf -\$1000} & {\bf -\$2000} & {\bf \$5000}\\ \hline \end{tabular}\]

So, the total profit over the year is $8,000, and then the average quarterly profit is $2000. The seasonal index (SI) for each quarter is then that quarter’s profit (or loss) divided by the average quarterly profit:

    \[ \begin{tabular} {| c | c | c | c |}\hline {\bf Summer SI} &{\bf Autumn SI} &{\bf Winter SI} &{\bf Spring SI} \\  \hline {\bf 3} & {\bf -0.5} & {\bf -1.0} & {\bf 2.5}\\ \hline \end{tabular}\]

Clearly this example is general, in the sense that in any scenario where the seasonal data are both positive and negative, some of the seasonal indices will be negative. So, the exam question is not merely technically wrong, with a contrived example raising issues: the question is wrong wrong.

Now, to be fair, this time the VCAA has a defense. It appears to be more common to apply seasonal indices in contexts where all the data are one sign, or to use absolute values to then consider magnitudes of deviations. It also appears that most or all examples Further students would have studied included only positive data.

So, yes, the VCAA (and the Australian Curriculum) don’t bother to clarify the definition or permitted contexts for seasonal indices. And yes, the definition in the standard textbook implicitly permits negative seasonal indices. And yes, by this definition the exam question is plain wrong. But, hopefully most students weren’t paying sufficient attention to realise that the VCAA weren’t paying sufficient attention, and so all is ok.

Well, the defense is something like that. The VCAA can work on the wording.


Three Apples + Two Oranges = Infinite Nonsense

The key findings of Australia’s 2016 National Drug Strategy Household Survey were released earlier this year, and they made for sobering reading. The NDSHS reported that over 15% of Australians had used illicit drugs in the previous year, including such drugs as cannabis, ice and heroin. Shocking, right?

Wrong. Of course.

We’re being silly in a way that the NDSHS reporting was not. Yes, the NDSHS reported that 15% had used illicit drugs at least once (including the possibility of exactly once) in the previous year, but NDSHS also emphasised the composition of that 15%. By far the most commonly used drug was cannabis, at about 10% of the population. Ice use was around 1%, and heroin didn’t register in the summary.

Illicit drug use is a serious problem, and a problem exacerbated by idiotic drug laws. Nothing can be learned, however, and nothing can be solved if one focuses upon a meaningless 15% multicategory. Whatever the specific threats or the reasonableness of concerns over the broad use of cannabis, such concerns pale in comparison to the problems of ice and heroin. The NDSHS makes no such categorical mistake. Unfortunately, there are plenty of clowns who do.

Last week, the Federal Ministers for Social Services and Human Services announced the location of a drug testing trial for job seekers who receive federal benefits. The ironically named Christian Porter and the perfectly named Alan Tudge announced that receipients would be tested “for illicit substances including ice (methamphetamine), ecstasy (MDMA) and marijuana (THC) … People who test positive to drug tests will continue to receive their welfare payment but 80 per cent of their payment will only be accessible through Income Management.” The plan is deliberately nasty and monumentally stupid, and it has been widely reported as such. For all the critical reporting, however, we could find no instance of the media noting the categorical lunacy of effectively equating the use of ice and ecstasy and THC.

Still, one should be fair to Porter and Tudge. They are undeniably dickheads, but Porter and Tudge are hardly exceptional. They are members of a very large group of thuggish, victim-blaming politicians, which includes Malcolm Turnbull, and Peter Dutton, and Adolf Hitler.

It is also notable that this kind of multicategory crap is only practised by social conservatives. It’s not like a nationwide survey on sexual harrassment and sexual assault in universities would ever couch the results in broadly defined categories in such a clouded and deceptive manner. Nope, not a chance.