Posts written by Thomas Lumley (2645)

This blog has a tendency to look pro-drug and pro-alcohol, because we react to media stories and these topics are currently popular for exaggeration and over-interpretation.  It’s a nice change to find some serious anti-drug research that the media are reporting pretty sensibly. (Otago Daily Times, Stuff, Herald, 3News, TVNZ)

Researchers in Dunedin, and their international collaborators, used data from the Dunedin Study, which recruited about 1000 babies born in a Dunedin hospital in the 12 months starting April 1972.  These babies, now adults, have given a lot of data up to medical science. In particular, they had a range of neuropsychological tests at age 13 and then again at age 38, and they were interviewed about cannabis use.   The headline results were about IQ, but the study looked at a relatively broad range of cognitive function tests, and also at self-reported problems with memory and attention.   Participants who smoked cannabis as teens had broadly worse cognitive function, by a margin that’s large enough to be a concern.  This didn’t show up in people who started smoking cannabis as adults.  The paper is in PNAS; the abstract is free.

What’s most distinctive about this study is the before-after comparisons.  You often see (and we have pointed out) studies that claim to have found “changes” when they only had one time point, and so really only looked at “differences”.  The Dunedin Study has changes from before cannabis use to after cannabis use.   Also, since the participants are the same age and all from Dunedin, they are less heterogeneous in many ways than participants in other research studies.

The differences in IQ are within the range of the Flynn effect, so it’s still possible that this is a social effect rather than a biochemical one, or that it’s caused by pre-existing differences in people’s interest in the sort of activities that increase IQ test scores, or something.  Better data, however, will be extremely difficult to collect, and treating these results as probably true seems sensible.  Presumably the adverse effects would also be found with synthetic cannabimimetics, but these are too recent for there to be data, and they have other problems not present with cannabis, such as the risk of acute overdose.

A number of the stories quote Ross Bell, of the Drug Foundation, being sensible:

Simply banning drugs such as cannabis and then thinking that was the problem solved would not work, Mr Bell said. “What we’re lacking in New Zealand is support for widespread, high-quality, well-constructed prevention messages targeted at younger people.”

Three sentences from a Herald story 

The Ministry of Transport study used blood samples taken from 453 drivers who caused crashes.

Of that group, 156 were found to be on drugs not administered by a medical professional

Drivers with more than the legal limit of alcohol in their system made up just over half of the 453 samples analysed.

Now try to fill in the blank  in the headline “Tests reveal most crash drivers                                “

If you said “were drunk”, your arithmetic is better than your sense of headlineworthiness, not a problem the Herald had.

A nice page by Christopher Healey, at North Carolina State University.  Among other things, it includes demonstrations of change blindness  and of preattentive perception:

For many years vision researchers have been investigating how the human visual system analyses images. An important initial result was the discovery of a limited set of visual properties that are detected very rapidly and accurately by the low-level visual system. These properties were initially called preattentive, since their detection seemed to precede focused attention. We now know that attention plays a critical role in what we see, even at this early stage of vision. The term preattentive continues to be used, however, since it conveys an intuitive notion of the speed and ease with which these properties are identified.

Typically, tasks that can be performed on large multi-element displays in less than 200 to 250 milliseconds (msec) are considered preattentive. Eye movements take at least 200 msec to initiate, and random locations of the elements in the display ensure that attention cannot be prefocused on any particular location, yet viewers report that these tasks can be completed with very little effort. This suggests that certain information in the display is processed in parallel by the low-level visual system.

The limits of preattentive perception are why you can’t usefully represent more than four or five groups of points in a scatterplot, no matter how creative you get with colours and symbols.

Stuff has a story pointing out that conspiracy theories go together:

If you think smoking doesn’t cause lung cancer, HIV doesn’t cause Aids or Nasa faked the Moon landing, you are also more likely to support free market economics and be sceptical about climate change.

An apparently stronger version of this was demonstrated in a study published in January this year,

A New York Times story about buying and selling favorable online reviews

Many of the 300 reviews he bought through GettingBookReviews were highly favorable, although it’s impossible to say whether this was because the reviewers genuinely liked the books, or because of their well-developed tendency toward approval, or some combination of the two.

[Update: XKCD]

XKCD (come on, you know XKCD): I can’t help but admire the audacity of the marketer who came up with the phrase “contains a clinically studied ingredient”

headsup (journalism blog): It’s more likely that the [Detroit Free Press] doesn’t understand that you’re supposed to do a little basic arithmetic before you talk about public opinion.

Andrew Gelman: graphs showing uncertainty in a fitted curve.

journalism.org.nz: a proposal for NZ public interest journalism funded by the public

BBC: Gathering evidence for the effects of exercise on depression is harder than you might expect.

Stuff is reporting on a software system called Volpara for improving breast-cancer screening, from Wellington company Matakina (and 3News had a similar story).  The software is designed to give more reliable and objective estimates of breast density, with less variation due to different equipment or different radiologists.  This should be helpful in screening younger women, either to improve detection of tumours or to reduce false positives, and the software seems to have been quite successful commercially.

Medical progress is pretty incremental: even the best ideas and techniques tend to have relatively small impact on health, so while it’s good to see NZ medical contributions being recognised, it’s also useful to put them in context. (more…)

The Herald says

One in three New Zealanders have been harmed by their own alcohol drinking, a survey shows.

The survey, published in the New Zealand Medical Journal, found 33.8 per cent of current drinkers reported they had been adversely affected in the past year.

This is actually true: there is such a survey, it is published in the NZMJ, and it’s a real survey (the response rate isn’t all one could wish, but it’s not out of line with other major health surveys).  The story goes on to say that the harm was reported more often by men.

The interesting point is that the current issue of NZMJ has two articles about frequency of harm from drinking.  The other one says, also based on real survey data,

The prevalence of self-reported harm from others’ drinking was higher than harm from own drinking (18% vs 12% in the past year) and was higher in women and young people.

And this second survey isn’t pro-alcohol in the slightest — for example, it quotes the proportion of criminal offenders who had been drinking as if it was the proportion of crimes resulting from alcohol.  This is a huge overestimate, precisely because there is too much drinking in NZ: the people who are drunk and arrested in the major cities tend to be just as drunk about three times a week on average (based on Massey Uni. research that we’ve mentioned before).

The timing of these articles is perhaps not a coincidence, but in fact there’s pretty wide consensus about the harms from drinking: the real disagreement is about the costs of various sorts of regulation.

At Simply Statistics, a set of stat/data related startups.

One that looks interesting for teaching and for data journalism purposes is Statwing, which is building a web-based pointy-clicky data analysis system, aiming to have good graphics and good text descriptions of the results.  This is the sort of project where the details will matter a lot — poking around at their demo there were a few things I was slightly unhappy about, but nothing devastatingly bad, so there is potential.

Those are the basic questions you need to ask when reading, for example, that “almost half of Australian workers would rather quit a job than deal with office tension.”

Unfortunately, in the case the media release doesn’t answer the first question, and its answer to the second and third questions is

The 2012 R U OK? Australian Workplace Relationships Survey was developed in partnership with the Centre for Corporate Health, one of Australia’s leading workplace mental health service providers. 

Now, workplace mental health is an important issue, and for all I know it may be seriously neglected in Australia (I last worked there in 1995, but it was in a medical research centre, which probably isn’t typical).   That doesn’t mean mental-health advocates get a free pass on data quality — good facts matter more for important issues than for, say, whether we should wear pyjamas in the street.

The Centre for Corporate Health isn’t that enthusiastic about telling us how the survey was done: you need to subscribe to their research news to get in, and even then all they say is

The survey was distributed via online and paper-based to people across a wide range of Australian workplaces including both public and private sectors with all industry groups being represented.  The survey was open for a four week period. The total number of survey respondents was 1554.

which is not encouraging either in what it says or what it omits.  As far as I can tell, there’s no reason to believe the numbers are good estimates of anything about Australian employees as a whole.

The reason that organizations do these surveys seems to largely be marketing: having things that look (from the right angle, in poor light) like statistics makes it easier to get media coverage.  It shouldn’t: if we’re covering West Island business or health news, “R U OK Day” is important enough to be in the papers without needing unconvincing numerical camouflage to get it past the gate.