Posts written by Thomas Lumley (1096)


Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient

April 11, 2014

Past performance no guarantee of future results

From the ACC

Julius Caesar was warned to beware the ‘Ides of March’. And perhaps Kiwis should take extra caution this Sunday.

That’s because April the 13th last year was the day on which the highest number of injuries happened during 2013.

Of course, April 13th wasn’t a Sunday last year. ACC helpfully give us the top five days last year for injuries

  • 13 April – 8,067 claims
  • 6 April – 8,024 claims
  • 11 May – 7,988 claims
  • 18 May – 7,757 claims
  • 8 June – 7,732 claims.

What do all these days have in common? Well, let’s just say that the ACC warning for Sunday April 13 may be a bit late.


The favourite never wins?

From Deadspin, an analysis of accuracy in 11 million tournament predictions (‘brackets’) for the US college basketball competition, and 53 predictions by experts



Stephen Pettigrew’s analysis shows the experts average more points than the general public (651681 vs 604.4). What he doesn’t point out explicitly is that picking the favourites, which corresponds to the big spike at 680 points, does rather better than the average expert.


April 10, 2014

Frittering away

Q: Did you see that “some generation Y foodies are spending up to $600 a week on gourmet produce such as seafood, cheeses, olives and cured hams.”

A: In the Herald? Yes.

Q: Is it true?

A: Slightly.

Q: Who are these people?

A: Well, for a start, they’re Australians

Q: Oh. How many is “some”

A: At least one.

Q: No, seriously, how many?

A: 1% of a the 18-34 subset of a sample of ‘over’ 1000. Here’s the full report

Q: How many is that?

A: Maybe three in the sample?

Q: Three people or three households?

A: A good question. They don’t say, though the average weekly food expenditure in their sample looks reasonably close to the national household average that they cite.

Q: How were the people sampled?

A: They don’t say.

Q: How many were Generation Y?

A: They don’t say

Q: How did they even define ‘gourmet food’? Or don’t they say that either?

A: Sadly, no.

Q: This report doesn’t seem to follow the code of practice you blogged about recently, does it?

A: That was just for political polls, and anyway this report is Australian.

Q: Is there anything else you want to complain about in the report?

A: If  you call it an “Inaugural” report you really can’t use it to conclude “Australians are becoming a more food savvy nation”.


April 9, 2014


Pie chart edition


I couldn’t possibly comment

Two excerpts from Gavin White’s  SAYit blog:

The flag

The specific question was “do you think the New Zealand flag should be changed?”.  The answers were:

  • Yes 19%
  • No 77%
  • Unsure 4%

And on conspiracy theories

…the moon landing:

  • 20% of us think that it’s likely that the United States of America government staged or faked the Apollo moon landing (5% very likely, 15% somewhat likely)
  • 74% think it’s unlikely
  • 6% are unsure or haven’t heard of this conspiracy before..

Busable Wellington

In response to a reader request, here are the same bus service maps as for Auckland. Again, click for big PDF files so you can zoom in.  I think these include non-bus transit, but I’m not completely sure.

Hours per day with at least six transit trips within 500m



Hours per day with at least 12 transit trips within 500m



In the Wellington region it seems that people who have any bus service have a useful amount.

The region as a whole has a smaller proportion of people with good public transport than Auckland region (7% in the top category, 60% in the bottom), but if we restrict to Wellington City there are 17% in the top category, 26% in the next, and only 30% in the lowest.


[Update: Here's a version with a 1km distance instead of 500m]


April 8, 2014

Asthma inhalers and diet: shorter, with more swearing

Ok, so the previous post is about Herald (Daily Mail) story on asthma research. As science reporting goes it’s no worse than usual for these Mail reprints. The reason for this second post is that I read the story again and thought about health reporting.

The story lead says

Eating fast food and consuming sugary drinks renders the most common asthma inhaler ineffective, a study warns.

 That is, the Herald is telling people their emergency asthma inhaler will not work if they eat certain foods. There’s no suggestion of what to do instead in an attack or who to call for help. Even if the claim were true, that would be irresponsible. When it’s just linkbait, it’s fscking appalling.

Overinterpreting diet and asthma

The Herald’s lead

Eating fast food and consuming sugary drinks renders the most common asthma inhaler ineffective, a study warns.

This is two studies. One looked at a ‘dietary inflammation index’ and whether people with higher values were more likely to be asthmatic. It did not look at inhaler effectiveness at all.  The dietary inflammation index does not measure ‘sugary drinks’; it treats all carbohydrate the same. It doesn’t directly measure fast food, though it does distinguish different types of fat.  Since the dietary inflammation index, according to the paper that proposed it, is relatively weakly associated with biological measures of inflammation,  the strong association seen in this study between the index and asthma may just mean that other factors are affecting both asthma and diet.

The other study looked at how salbutamol, the active ingredient of the Ventolin inhaler, was absorbed in samples of lung tissue in the lab. The amount of polyunsaturated fat affected absorption — but that wasn’t dietary fat and there weren’t any inhalers, or any sugary drinks.  Further research might show this translates into real differences in inhaler effectiveness, or it might not.

So, while there is actual science described in the article, there is almost no support for the lead. No inhalers, no sugary drinks, no fast food.

You might also wonder why the Herald is getting a comment from Asthma UK for research presented at a conference of the Thoracic Society of Australia and New Zealand. Or not.

Busable Auckland

Bus commuter services can be very useful in reducing traffic and parking congestion in the city center, but reducing the average number of cars per household requires buses that are available all the time. I used the Auckland Transport bus schedule data and the new StatsNZ meshblock data and boundary files

Here’s a map of Auckland showing how many hours per day (on average) there are at least six bus trips per hour stopping within 500m of each meshblock (actually, within 500m of the ‘label point’ for the meshblock).

On a single road, six trips per hour is one trip in each direction every twenty minutes. The dark purple area has this level of service at least 16 hours a day on average. (Click for the honking great PDF version.)


For twelve trips per hour (eg, one every twenty minutes on two different routes) the area shrinks a lot


The reason for using meshblocks in the map is that we can merge the bus files with the census files. For example, for Auckland as a whole, 50% of the population is in the grey busless emptiness, 17% in the 8-16 hour tolerable zone, and 12% in the pretty reasonable 16+ hour zone.   People of Maori descent are more likely to be unbused (60%) and less likely to be well bused (8%), as are people over 65 (60% in the lowest category, 9% in the highest).

Recent (<10 years) migrants like transit: 18% of us are in the good bus category and only 40% in the busless category.

Another phony poll?

The flipside of failing to question numbers is assuming that anyone who does question them is correct (from US journalism/editing blog headsup)

Hence, even though in this case it’s an out-and-out lie, “another phony poll” doesn’t just resonate around the echo chamber; it puts a bug in the grownup world’s ear too. It looks like what journalism is supposed to be doing. Polls are suspect anyway (being based on data, which is inferior to gut feelings and news sense), and catching a fake one is a triumph of  the common man over the spinmeisters