April 19, 2014

Seeing no evil

The Herald story “Uni cheats: hundreds punished” is a pretty good example of using actual data to combat the ‘false balance’ problem in journalism.  The story notes the huge variation in official proceedings for cheating between NZ universities — nearly half the cases are at Waikato — and rightly highlights it as “suggesting differences in what institutions consider cheating, and how they target and record it.”

As the story doesn’t point out, with 540 cases out of more than 400000 tertiary students it’s pretty clear cheating is underreported everywhere. That’s hardly surprising, given the costs and benefits to staff for following it up. If it wasn’t for the irrational rage cheating arouses in academics, it would be perfectly safe.


There’s nothing to it

From the Herald, in a good article about the Australian report on homeopathy

Auckland homeopath Suzanne Hansen said the treatments could not be measured in the same way medical treatments were.

“When you research it against a medical paradigm it will fail because you treat in a completely different way.”

This is probably true, but it’s a major concession that should be noted for the record.

The`medical paradigm’ of randomised controlled trials doesn’t need treatment to be the same for each person, it doesn’t need the benefits to be the same for each person, it doesn’t need the biological mechanism to be known or even plausible. All they need is that you can identify some group of people and a way of measuring success so that getting your treatment is better on average for your chosen group of people with your chosen way of defining ‘better’.  This isn’t just about homeopathy —  the whole field of personalised genomic medicine is based on individualising treatment, and this doesn’t introduce any difficulties for the medical paradigm.

If an intervention can’t beat fake pills that do nothing, on its choice of patient group and outcome measurement, it will fail when you `research it against a medical paradigm.’  If you’re fine with that, you should be fine with not using any advertising terms that suggest the intervention has non-placebo benefits.

April 18, 2014


Cannabis and Radio Yerevan

Radio Yerevan jokes were a thing in the Soviet Union days

The Armenian Radio was asked: “Is it true that in Moscow, Mercedes cars are being given to citizens?”

The Armenian Radio answers: “Yes, but it is not Moscow but Leningrad, not Mercedes but Ladas, and not given to but stolen from.” 

From the Herald, yesterday

People who had only used cannabis once or twice a week for a matter of months were found to have changes in the brain that govern emotion, motivation and addiction.

Here’s the research paper, which you won’t be able to access, so I’ll summarisse

Firstly, no-one in the study was found to have ‘changes’ in the brain: the participants got only one brain scan, so the research didn’t even look at changes. It found differences between cannabis users and non-users. There’s nothing even slightly surprising about the possibility that people who end up as regular users of an illegal drug might have started off with brain differences.

There were 20 cannabis users in the study, who smoked an average of 11 joints per week, and had been smoking cannabis for an average of six years. Here’s the data for one of their findings, on ‘gray matter density in the nucleus accumbens’



The dots at zero are the controls, the other dots are the cannabis users. There certainly aren’t many who use cannabis only once or twice per week. It’s hard to tell whether there’s really anyone who has only used cannabis for a few months,  because the research paper only reports the mean and standard deviation for duration of use.

So, the main point of the story in the paragraph quoted above is completely unsupported by the research.  This one isn’t entirely the fault of the media, since the researchers were pushing hard to exactly this sort of unsupported claim into the papers.  Still, you might have hoped someone they talked to would have matched up the claims to the research..



April 17, 2014

This is not a map



This is not a map. The Asian population of the US is not confined to Maine and northern Washington, and residents of the Dakotas are not primarily Black and Hispanic. It’s a stacked line plot, which has been cut out to fit the map outline, just like you might do in kindergarten. (via Flowing Data)

Here’s the real thing, from Pew Research.


April 16, 2014

NRL Predictions for Round 7

Team Ratings for Round 7

The basic method is described on my Department home page. I have made some changes to the methodology this year, including shrinking the ratings between seasons.

Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Current Rating Rating at Season Start Difference
Roosters 7.55 12.35 -4.80
Rabbitohs 6.40 5.82 0.60
Bulldogs 6.24 2.46 3.80
Sea Eagles 5.95 9.10 -3.10
Cowboys 2.89 6.01 -3.10
Knights 2.86 5.23 -2.40
Storm 2.24 7.64 -5.40
Titans 0.26 1.45 -1.20
Broncos -1.83 -4.69 2.90
Sharks -2.75 2.32 -5.10
Panthers -2.92 -2.48 -0.40
Wests Tigers -4.13 -11.26 7.10
Warriors -4.40 -0.72 -3.70
Raiders -5.45 -8.99 3.50
Dragons -5.47 -7.57 2.10
Eels -9.20 -18.45 9.30


Performance So Far

So far there have been 48 matches played, 24 of which were correctly predicted, a success rate of 50%.

Here are the predictions for last week’s games.

Game Date Score Prediction Correct
1 Panthers vs. Rabbitohs Apr 11 2 – 18 -2.30 TRUE
2 Titans vs. Broncos Apr 11 12 – 8 7.30 TRUE
3 Raiders vs. Knights Apr 12 12 – 26 -1.50 TRUE
4 Eels vs. Roosters Apr 12 14 – 12 -15.40 FALSE
5 Wests Tigers vs. Cowboys Apr 12 16 – 4 -5.70 FALSE
6 Warriors vs. Bulldogs Apr 13 20 – 21 -7.40 TRUE
7 Sea Eagles vs. Sharks Apr 13 24 – 4 11.60 TRUE
8 Storm vs. Dragons Apr 14 28 – 24 14.10 TRUE


Predictions for Round 7

Here are the predictions for Round 7. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Game Date Winner Prediction
1 Rabbitohs vs. Bulldogs Apr 18 Rabbitohs 4.70
2 Knights vs. Broncos Apr 18 Knights 9.20
3 Sea Eagles vs. Cowboys Apr 18 Sea Eagles 7.60
4 Dragons vs. Warriors Apr 19 Dragons 3.40
5 Sharks vs. Roosters Apr 19 Roosters -5.80
6 Raiders vs. Storm Apr 20 Storm -3.20
7 Eels vs. Wests Tigers Apr 21 Wests Tigers -0.60
8 Panthers vs. Titans Apr 21 Panthers 1.30


Super 15 Predictions for Round 10

Team Ratings for Round 10

The basic method is described on my Department home page. I have made some changes to the methodology this year, including shrinking the ratings between seasons.

Here are the team ratings prior to this week’s games, along with the ratings at the start of the season.

Current Rating Rating at Season Start Difference
Sharks 7.52 4.57 3.00
Crusaders 7.20 8.80 -1.60
Chiefs 4.10 4.38 -0.30
Brumbies 4.06 4.12 -0.10
Waratahs 3.18 1.67 1.50
Bulls 2.88 4.87 -2.00
Stormers 0.66 4.38 -3.70
Hurricanes 0.51 -1.44 1.90
Reds -0.84 0.58 -1.40
Blues -1.33 -1.92 0.60
Force -2.44 -5.37 2.90
Highlanders -3.83 -4.48 0.70
Cheetahs -4.29 0.12 -4.40
Rebels -5.18 -6.36 1.20
Lions -5.22 -6.93 1.70


Performance So Far

So far there have been 55 matches played, 34 of which were correctly predicted, a success rate of 61.8%.

Here are the predictions for last week’s games.

Game Date Score Prediction Correct
1 Highlanders vs. Bulls Apr 11 27 – 20 -4.10 FALSE
2 Reds vs. Brumbies Apr 11 20 – 23 -2.30 TRUE
3 Chiefs vs. Rebels Apr 12 22 – 16 14.40 TRUE
4 Force vs. Waratahs Apr 12 28 – 16 -5.20 FALSE
5 Cheetahs vs. Crusaders Apr 12 31 – 52 -5.60 TRUE
6 Lions vs. Sharks Apr 12 12 – 25 -9.80 TRUE


Predictions for Round 10

Here are the predictions for Round 10. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

Game Date Winner Prediction
1 Hurricanes vs. Blues Apr 18 Hurricanes 4.30
2 Rebels vs. Force Apr 18 Force -0.20
3 Chiefs vs. Crusaders Apr 19 Crusaders -0.60
4 Waratahs vs. Bulls Apr 19 Waratahs 4.30
5 Sharks vs. Cheetahs Apr 19 Sharks 14.30
6 Stormers vs. Lions Apr 19 Stormers 8.40


Ways not to use the Global Drug Use Survey

We learned previously from Stuff and the Global Drug Use Survey that 22% of New Zealanders have used synthetic cannabis. Today

Results from this year’s Global Drug Survey, conducted in partnership with Fairfax Media, found almost 4 per cent of synthetic cannabis users sought emergency medical treatment. More than a quarter of those were admitted to hospital.

It simply cannot be true that 4% of 22% of the country has sought emergency treatment after using synthetic cannabis. Even restricting to adults, that’s 30,000 people, with more 7,500 admitted to hospital. In the most recent year for which I can find data (2010-11, when the drugs were more widely available than now) there were 672,000 publicly funded hospital admissions for all causes, and of those, only 896 were for cause categories X41 & X42, which would include all synthetic cannabis cases plus many others.

[update: fixed typo in numbers]

April 14, 2014

What do we learn from the Global Drug Use Survey?



That’s the online summary at Stuff.  When you point at one of the bubbles it jumps out at you and tells you what drug it is. The bubbles make it relatively hard to compare non-adjacent numbers, especially as you can only see the name of one at a time. It’s not even that easy to compare adjacent bubbles, eg, the two at the lower right, which differ by more than two percentage points.

More importantly, this is the least useful data from the survey.  Because it’s a voluntary, self-selected online sample, we’d expect the crude proportions to be biased, probably with more drug use in the sample than the population. To the extent that we can tell, this seems to have happened: the proportion of past-year smokers is 33.5% compared to the Census estimate of 15% active smokers.  It’s logically possible for both of these to be correct, but I don’t really believe it.  The reports of cannabis use are much higher than the (admittedly out of date) NZ Alcohol and Drug Use Survey.  For this sort of data, the forthcoming drug-use section of the NZ Health Survey is likely to be more representative.

Where the Global Drug Use Survey will be valuable is in detail about things like side-effects, attempts to quit, strategies people use for harm reduction. That sort of information isn’t captured by the NZ Health Survey, and presumably it is still being processed and analysed.  Some of the relative information might be useful, too: for example, synthetic cannabis is much less popular than the real thing, with past-year use nearly five times lower.

Peak car?

From the Herald, quoting the chief executive of Z Energy:

“People are doing online shopping and Skyping granny rather than making the fortnightly visit.”

A 1 per cent improvement in broadband connectivity is estimated to cause a drop of 200 million litres a year in national fuel demand, more than the impact of GDP growth, population, fleet turnover, vehicle efficiency and the petrol price.

The first question here is on units. For broadband, it’s fuel demand per 1% of connections, but what are the units for the others?

There’s a bit more detail in this set of slides, including this picture, where the orange bar shows the estimated effect of an increase in the factor and the yellow bar shows the estimated effect of the same decrease.

So if we believe these numbers, a 1% point increase in broadband has slightly larger impact than a 1% increase in GDP and about twice the impact of a 1% increase in population.

For this model to be useful in prediction, which is what Z Energy presumably made it for, there’s no need that these statistical associations are causal. It’s only necessary that they persist at roughly the same strength through the period of the forecast.  The associations can’t really be true under serious extrapolation. For example, reducing broadband coverage from the current roughly 80% of households to zero would probably not cause transport fuel use to rise by 16 billion litres — ie, more than triple. Similarly, it can’t really be true that the impact nominal petrol prices is independent of inflation or income trends. For prediction this doesn’t necessarily matter, but for interpreting causes it does.

The actual prediction impact of broadband depends on how much it will increase. It turns out that the model says the reduction due to broadband plus the reduction due to increasing petrol prices approximately cancels out the increase due to increasing GDP. So, in fact, in the Z Energy model, broadband is less important than GDP growth. The model ends up predicting that per-capita travel will be roughly constant,  that total travel will increase with population, and that fuel efficiency will increase.

So, is the broadband association causal? It easily could be. There’s evidence from other countries of a reduction in driving that can’t entirely be attributed to the Great Recession. This is especially true among young people, with more socialising electronically. Telecommuting probably plays a role, too. I’m not convinced that online shopping has had a big impact on car trips in NZ, but it could have.  On the other hand, there huge uncertainty in the size of the effect — not just statistical uncertainty based on the data, but uncertainty about what’s fundamentally going on.

Finally, one depressing, but probably accurate, feature of the predictions is that they assume we still won’t be doing anything about climate change by 2018.