Stats Chat

Stuff:

The Dow Jones industrial average was set for its worst start to a year since 1932 as stock markets tanked after weak Chinese economic data reignited fears of a global slowdown.

Herald

The U.S. blue-chip index tumbled toward its worst start to a year since 1932, while banks and technology shares led the Standard & Poor’s 500 Index lower.

That’s presumably true, and bad news, but “start to a year” is doing a lot of work. Here’s the past year’s Dow:

Unless there’s something especially predictive about the first trading day of the year, this is no worse than some quite recent days. And, the Herald says further down in the story

S&P Dow Jones Indices data indicate the first day of trading has no predictive power for the rest of the year. The index ends the year in the same direction it takes on the opening day 50.6 percent of the time, the data show. The first month of the year has proved more telling — the gauge’s return in January determines its direction for the year 72.4 percent of the time.

Lateral is a relatively new science magazine produced by a bunch of mostly Australian, mostly young over-achievers. Their sixth issue is about Data.

And I think I’ve mentioned STAT, the new health and medicine (and not statistics) news site, but it’s worth mentioning again.

There’s another Herald story about incomes based on job ads at Seek.co.nz.

Data released by job search company Seek shows outside of consultancy work roles linked to the building industry were paid the most last year and were some of the few sectors to see decent pay rises.

The average salary for the construction industry is now $94,580, a boost of 5 per cent on 2013 while engineers earned an average of $92,595

Stats NZ data isn’t quite as up to date, but the NZ Income Survey, in June, found average weekly earnings in the construction industry to be $1096 (go here, and select ‘Construction’ from ‘Industry’), up about 1% from 2013 (though 5% from 2014). If you assume a full-time job with holidays, that’s $57,000 per year.

I don’t know how much of this is due to different definitions and how much is due to the Seek jobs being non-representative, but it’s possible for anyone who really cares to find out exactly what the Stats NZ number means, and that’s not true for the Seek number.

• True-colour photos of Earth from space — why it’s not as simple as that, by Emily Lakadawalla
• “The ranks of the wealthy grew in 2013, with 1.38 million households making it into the Top 1%, up from 1.36 million a year earlier.”  From CNN Money (who should understand percentages) via Felix Salmon
• Translation of international new reprints matters. The Herald had a headline “North Pole now 50 degrees above normal“, which was 50 Fahrenheit degrees. In NZ terms, just under 30 Celsius degrees.
• From the Christmas edition of BMJ, famous for its joke (but accurate) articles, a functional MRI imaging study comparing people who do and don’t celebrate Christmas, found differences in brain activation. These differences could be argued as a biological “Christmas Spirit” locus, plausible in terms of previous research.  The point, obviously, is how easy it is to overinterpret such things.
• A group in the US did a randomised, blinded trial to see what effect faster internet speeds would have: they raised the bandwidth for 1500 people from 105Mbps to 250Mbps without telling them. People used the internet more, and it wasn’t primarily the heaviest users whose use increaseed.

Every summer, the Department of Statistics offers scholarships to a number of students so they can work with staff on real-world projects. Jale, right, is exploring evolutionary relationships with Dr Steffen Klaere. Jale explains: 

“To understand the evolution of life on earth, we need to make inferences about evolutionary events leading up to the diversity of life we see now. My project is about phylogenetic inference – the set of bioinformatic tools for estimating evolutionary relationships between different species, or taxa.

“The set of taxa and their divergence is usually represented by a DNA-sequence alignment. The basic assumption is that each sequence represents a taxon and the evolutionary divergence between species is identified by differences in the respective sequences. The relatedness of taxa is then represented by a phylogenetic tree, where closely-related species are identified with leaves that are close together in the tree.

“In my project, I will investigate different statistics to assess the fitness between model and data. In particular, I am interested in identifying sites in the alignment that are not well represented by the model.

“It has been established that alignments for species with an old, most common recent ancestor will have more sites that will not inform the phylogenetic hypothesis. Such sites are often called saturated sites due to the assumption that they accumulated a lot of mutations over time. It has been hypothesised that such sites can lead to systematic error in the inference.

“That is why we want to identify influential outliers and mask them for the inference. Statistics like observed variability (OV) distances have been proposed, which are easier to compute, but tend to overestimate the number of saturated sites.

“My task will be to investigate methods that propose to identify saturated sites and test them on datasets known to suffer from systematic error. In particular, I want to test the utility of combining different statistics to address such problems.

“I am from Duelmen in Germany, but I have lived and studied in Greifswald for the past three years. I have just completed my Bachelor of Science in Biomathematics at the University of Greifswald. After this research project, I would like to pursue my Master of Science in Statistics at the University of Dortmund, also in Germany.

“What I like about statistics is that it has a wide area of use and deals with diverse topics in biology, medical science and economics. Furthermore, I like that statistics deals with all aspects of data, including the planning of data collection in terms of the design of surveys and experiments. You can work together with people from different institutions and you benefit from the different knowledge they bring.

“As I have some free time over Christmas and New Year, I am thinking about travelling to the Bay of Islands to spend some time at the beaches and enjoy the beautiful countryside. After the research project, I will also have some time to travel to the South Island before flying back to Germany.”

From the Herald (from the Telegraph)

Using e-cigarettes is no safer than smoking tobacco with nicotine, scientists warned after finding the vapour damages DNA and could cause cancer.

Smoking tobacco is right up near the top of cancer risks that are easy to acquire, both in terms of how big the risk is and in terms of how strong the evidence is.

[There was some stuff here that was right as to the story in the Herald but wrong about the actual research paper, so I got rid of it. Some of the tests in the research paper used real cigarette smoke, and it was worse but not dramatically worse than the e-cig smoke]

The press release is a bit more responsibly written than the story. It describes some of the limitations of the lab tests, and makes it clear that the “no safer than smoking” is an opinion, not a finding. It also gets the journal name right (Oral Oncology) and links to the research paper.

It’s worth quoting the conclusion section from the paper. Here the researchers are writing for other people who understand the issues and whose opinion matters. I’ve deleted one sentence that’s technical stuff basically saying “we saw DNA damage and cell death”

The most obvious thing in the StatsChat logs: the Rugby World Cup:

Also, back last January, there was a study on the relationship between cell divisions and cancer risk across human tissues. The popular misinterpretations of the research — “cancer is mostly bad luck” — led to our most popular post ever.

The question of what it means for something to be a Group I carcinogen gets us a lot of low-level traffic, but interest peaked after the IARC report on red meat and processed meat.

Posts on the risk posed by foreign drivers were popular early in the year. In July, though, they were displaced by foreign-sounding home buyers.

I wrote about the largest human randomised controlled trial of mānuka honey to prevent illness, when it was reported in June. It was done by kids at a London primary school. They didn’t find a benefit.

Finally, there’s a steady trickle of people interested in the mathematics of the lottery, presumably in the mistaken hope that we’ll tell them how to beat the martingale optional sampling theorem.

Q: Did you see there’s a designer Kiwi cider that makers claim could combat ageing?

A: <sigh> Yes.

Q: And do the makers claim that?

A: Pretty much: “I never feel guilty about the drinks I consume, but I know that’s not the case for everyone. Knowing I’ve played a part in unlocking the secret to eternal youth is a career highlight.”

Q: Wow.

A: I know, right?

Q: What is his secret to eternal youth?

A: Cascara

Q: I think I’ve heard of ….  wait,  isn’t that a laxative?

A: They’re probably hoping their customers are too young to think of that. This is a completely different plant, just the same Spanish word.

Q: Ah. I see. Coffee berry husks. They used to just throw it away?

A: Well, in Yemen it’s a traditional beverage, and they drink it in Bolivia, too. And everywhere it at least gets recycled as compost. But from the viewpoint of the superfood industry they threw it away.

Q: Are they more specific about the whole eternal youth thing?

A: The Herald didn’t quote them, but yes, the company blog says “brewers added the berry’s purported health benefits that include preventing premature aging, heart disease, high cholesterol, asthma and Alzheimer’s to the cider profile. ”

Q: What research do they cite supporting this?

A: Very droll.

Q: No, seriously, don’t they say anything?

A: No.

Q: Is there research?

A: Well, it’s got antioxidants.

Q: Doesn’t everything?

A: And it seems to stimulate the release of brain-derived neurotrophic factor.

Q: In mice?

A: No, in people. Five of them.

Q: And is this brain-derived neuro-factor thing good?

A: With a name like that, how can it not be? Having some of it is important for lots of things.

Q: Some of it?

A: Doesn’t seem to be any real evidence that releasing more is good.

Q: But releasing it is unusual?

A: Caffeine does.

Q: Oh. Doesn’t sound like much support for the claims.

A: That’s what Dr Paul Jarrett, said — the expert the Herald asked.

Q: How about Yemen? Does this coffee berry stuff make them healthier there?

A: You couldn’t really tell, given what a mess the social and healthcare systems are in, not to mention the occasional drone bombings.

Q: Well, that was depressing

A: Ok. Back to the cider, then.

Q: Right. So, you’re not going drink it?

A: Well, not for the eternal youth. The flavour combination sounds attractive if I can hold my nose over the marketing.

Q: And maybe find a charity that does useful things in places like Yemen and Bolivia

A: Yeah.

• Di Cook on improving graphical displays of cricketers’ careers.

• Graphics actually: automatically drawing relationships between characters in Love Actually based on the scenes they occur in.

• Remake of a graph showing changes in global poverty.

• Quartz has a list of the most-misleading graphs of 2015

• Hyde Park (in London (the one in England)) was tracking visitors via their mobile phones. Not actually harmful, but maybe not a good look if you want people to trust you.

• The Herald (and to a lesser extent Stuff) has started reporting number of crashes as well as number of fatalities on the roads. Probably a good thing.

• Health News Review writes about spin in medical news reporting.

• “But now scientists may finally have the answer. They have found that if a food is labelled as a healthy option we tend to eat more of it” From the Mail via the Herald. In fact, the research did not compare the consumption of food with healthy labels to food without.

• The frequency of `positive’ words such as “favorable”, “encouraging”, “supportive” has been increasing in medical research abstracts. In less favourable reporting, Language Log points out that abstracts have gotten longer, and is not entirely supportive about deducing attitude from single words.

• Also from Language Log: the `traditional’ `grammar rule’ that “which” doesn’t occur in restrictive relative clauses first became popular roughly when The Beatles did.
  

For when stories we’ve commented on get recycled as ‘best of’

• Herald: Has an 8-year-old girl found a cancer cure?. StatsChat: Probably not.