Dance With Chance blog

Testing, testing, testing... is it necessary?

noreply@blogger.com (The authors) — Fri, 19 Jun 2009 11:30:00 +0000

The recent publication of two large-scale studies of prostate screening in the US and Europe attracted our attention. After all, (a) this was something we wrote about in our book, and (b) we are - all three of us - men of a certain age.

In the US study, 38,343 men received annual PSA (Prostate Specific Antigen) blood tests while 38,350 men were assigned to a control group. After 7 years of follow-up, the incidence of prostate cancer death per 10,000 person-years was 2% in the screening group - that is 50 deaths. In the control group it was 1.7%, or 44 deaths. Not particularly significant then.

The European study involved some 180,000 men. This time, however, the screening group was billed as having a 20.7% improvement in the survival rate over the control group. This seems impressive and a good reason for continuing with PSA screening - in Europe at least.

So are European doctors better at screening than their US colleagues? Or is it just that numbers can be deceptive?

There were 214 deaths from prostate cancer in the screening group and 326 in the control group. This means 112 fewer deaths in the screening group, or a relative improvement over the screening group of 20.7% - found by dividing 112 by 540 (540 = 214 + 326).

In absolute terms, however, the improvement is much less impressive, since there were 72,890 subjects in the screening group and 89,353 in the control group. This means 7.1 fewer deaths in the screening group than in the control group, for every 10,000 people. If we consider that the study lasted some nine years, this is a tiny improvement of less than one person per year.

So should society continue this kind of testing? The answer should depend on the costs as well as the benefits. The authors of the European prostate cancer screening study summarize them as follows: 1,410 men would need to be screened and 48 additional cases of prostate cancer treated to prevent one death from prostate cancer.

In addition, they say: of the men who underwent biopsy for an elevated PSA value, 13,308 (75.9%) had a false positive result. And the authors of the USA study talk about 68 complications (per 10,000 screened men) caused by the diagnostic process, including bleeding, clot formation, and urinary difficulties while the treatment complications, generally more serious, included infection, incontinence, impotence, and other disorders.

On a personal level, where does this leave the three of us? First, there's the cost and inconvenience of having a blood test, which we can probably dismiss as insignificant. Second, there's the possibility of being one of the 47 treated unnecessarily to save one life. That's a bit more worrying. Third, there's the chance of being one of the 68 in 10,000 who get serious side affects from the diagnostic process. Now, that really is starting to get serious. Finally, there's the unpleasantness of the diagnostic process itself. Those 13,308 American men who underwent the biopsy by mistake not only had a thoroughly uncomfortable experience, but probably also had a traumatic time until the PSA test was proved wrong. At this point, we're feeling ill just thinking about it.

In other words, there's a lot of uncertainty, even for the three of us with our strong statistical background. One thing is sure, though. To screen or not to screen is a personal decision that we won't be delegating to our doctors. We are, however, interested in what our readers think. What would you do?

Putting some randomness into the national economy

noreply@blogger.com (The authors) — Fri, 05 Jun 2009 13:12:00 +0000

In these tough global economic times, it's nice to see national economies harnessing the power of luck. The Greek finance ministry recently announced proposals to let people off paying their income taxes if their personal code matches the winning numbers in the national lottery. Apparently the idea is modeled on a similar scheme in India, which has had some success.

But most of all, it's inspired by estimates that the "unofficial" economy in Greece is worth 28% of the official GDP. If correct, that adds up to an awful lot of tax evasion... which of course is one of the most risky financial strategies of them all.

Will this novel proposal turn Greece into a nation of enthusiastic tax payers? We doubt it. But if the law is passed and it works, the irony would be satisfying. After all, it's become a tradition for countries to use national lotteries to avoid raising taxes. In the UK, for example, the national lottery has been a nice little earner for the arts and sports? but arguably even more so for the government, which doesn't have to dig so deep into the national purse to fund these important aspects of society.

The next logical step would be to introduce ring-fenced lotteries for more fundamental elements of national life, such as health or education. In fact that's not such a far-fetched idea after all. Lottery bonds have been commonly used in European countries (such as Belgium and Sweden) to raise government money. They vary from country to country, but the principle is the same. While some bonds earn modest or even zero interest, others are drawn from a metaphorical hat to receive much higher returns. It's a lucrative investment for the winners, while the losers don't lose any money at all.

And, with soaring national debts throughout the world, who knows how many governments will need to introduce innovations of this kind simply to sell off hard-to-shift bonds?

Meanwhile back in Greece, there's a new financial craze. It's called bank robbery. It seems that there are currently four or five of them every day, often carried out by ordinary citizens with previously unblemished records. Maxed-out credit cards and huge personal debts are often cited as mitigating circumstances.

Is there also an element of revenge for the global financial crisis, we wonder? And are these ill-gotten gains counted in the government's calculations about undeclared earnings? We have no answers to these questions. But we're pretty sure bank robbery is even riskier than tax evasion.

It's not easy to be random

noreply@blogger.com (The authors) — Wed, 27 May 2009 09:50:00 +0000

Our web master rang us the other day. She said that she'd selected the winners for our prize draw, which closed in early May. But there was a slight problem. "I didn't add any of that small print about friends and family not being allowed to enter," she explained, "as I was so confident of the random nature of my process. But, the thing is, three of the winners are my friends. What should I do?"

The process in fact consisted of assigning a unique number to each of the 615 entrants and then using a reputable random number generator on the web to select 13 winning numbers. So far, so random.

But, it turns out, computers aren't very good at being random. They're designed to programme anything but chance outcomes. That's why the best random number generators, like the one our webmaster used are highly sophisticated methods, based on real-world phenomena like variations in atmospheric noise, rather than computer algorithms that rely on pseudo-random computer algorithms. You can try it for yourself at www.random.org.

Having said that, there's no way the webmaster's computer could have influenced the machine in Dublin which was ultimately drawing the numbers for us. In fact, it was a much more random process than any other method we could think of.

For example, if you put pieces of paper in a hat - or similar - they rarely get mixed up properly. This is why more Americans with birthdays towards the end of the year got called up for the Vietnam War in 1969 than those with birthdays at the beginning of the year. That was a real life-or-death lottery, which eventually resulted in a legal challenge in the US courts. It really puts winning a book or not in perspective (even if it is a very good book).

Anyway, after a brief consultation, we discovered that none of our own friends were in the list of winners, even though several of them had entered. And our webmaster didn?t know many of the 602 losers. Overall, then, it was a surprising result, but not that surprising. And to redraw the numbers really would be to bias the results.

That's why the original list of winners stands. So what if a perfectly random process threw up some freak results? It's what happens when you dance with chance. At least, that's a better conclusion than the alternative... which is that our webmaster has a lot more friends than we do.

Not so much dancing as globetrotting

noreply@blogger.com (The authors) — Thu, 14 May 2009 10:25:00 +0000

We never said making luck work for you would be easy. In fact, in our own case, it's been pretty strenuous for the last few weeks. You see, in order to improve the book's chances of success we've been crisscrossing the globe relentlessly. After our big launch event in France (see earlier blog), Anil set off for China and Australia, while Robin toured the USA. If you were in any of these places, you might even have glimpsed us on television (Anil on Sky Australia and Robin on ABC News). Or perhaps you heard us podcasting globally from our various far flung locations.

But statistically, we know that it's most likely you missed us. So you'll find some links to our various media activities below.

And if you're wondering where Spyros is, he's philosophizing back home in Greece. Someone has to do the thinking after all.

Thank goodness there are three of us. The ability to be in three places at once just has to improve our chances of turning the book into a bestseller. Well, that's the theory anyway. For now, we'll just settle for some sleep.

Robin on U.S. national television
ABC News clip

Robin on the West Coast
A lovely 'View from the Bay'

Anil in Singapore

Robin on the radio
Total Picture interview (but no pictures)

Robin on the Invisible Hand podcast
The Invisible Hand (yes, it's invisible)

Robin on US web radio show McLoughlin at Work
Download a preview

We need a health warning for forecasts

noreply@blogger.com (The authors) — Mon, 11 May 2009 14:46:00 +0000

Throughout the current economic crisis, people have been talking about better regulation for the banking sector. But what about better regulation for the forecasting industry?

Across history, the human race has prospered by predicting. We've learned to forecast many important physical phenomena with stunning accuracy: sunrise and sunset, times and heights of tides, even - on a good day - when scheduled flights will arrive.

But in most other circumstances, we get it woefully wrong - as we may have said before.

Obviously, we can't stop people making predictions or even wanting to believe them. It's become pretty evident over the centuries that this is just human nature. We may have mentioned that before too. Only last week, in fact.

However - and here's the new bit of our thinking - we'd like to suggest a kind of 'code of conduct' for forecasters - a bit like the caveats used by financial advisers, but going a somewhat further.

In general, we advocate holding forecasters accountable. We'd even like to see them given and star ratings based on their past performances.

But more realistically, we propose that all published forecasts should be accompanied by the following information:

* A clear list of all assumptions made.
* A broad explanation of how the forecast will change if one or more of these assumptions turns out to be wrong.
* In addition to the main forecast, a range of values showing the uncertainty involved - together with an indication of the likelihood that even these ranges might be exceeded.
* For organizations issuing regular forecasts, a statement of their past accuracy, using simple benchmarks and even comparisons with simple alternative forecasting models.
* A reminder of the dangers of believing in inaccurate forecasts.

The last of these shouldn't be in its current small-print form but in big letters, like the health warnings on cigarette packets. In the end, the comparison with smoking seems particularly apt. As a human race, we need to kick our dangerous forecasting habit, before it inflicts further damage on our economic health.

Dancing with the Black Swan

noreply@blogger.com (The authors) — Tue, 05 May 2009 10:05:00 +0000

We recently met up with our friend and fellow believer in chance, Nassim Nicholas Taleb, author of Fooled by Randomness and The Black Swan. You can see him interviewing us below, courtesy of the wonderful INSEAD Knowledge website, where you'll find lots more food for thought.

Prediction in a crisis

noreply@blogger.com (The authors) — Fri, 01 May 2009 08:29:00 +0000

One of our first reviews expressed disappointment that the book was in production when the full horror of the economic crisis dawned. The reviewer wished that he'd been able to read our thoughts about the global subprime fall-out.

It was very sweet of him to say that. But we fear we'd have become quite boring. After all, if ever we needed proof of our inability to make accurate socio-economic predictions, it's the current economic crisis. Not only did the world fail to see it coming; the downturn is itself the result of some very dodgy forecasting.

So, rather than become boring, we'll just offer a single small piece of advice to all of us. Now it might sound a bit paradoxical, but we don't advocate giving up trying to predict altogether, crisis or not. There are two main reasons.

The first is that to predict is human: we just can't help wanting to see the future. The second is that, by keeping a record of our predictions, we can learn from our errors.

And perhaps the biggest error of all is to let our forecasting imaginations be dominated by the immediate past. At the end of the 1980s, for example, there was a flurry of books predicting depressions, while at the end of the 1990s the mood changed to one of euphoria.

And it's happening all over again. In January 2009, a book came out with the title: The Great Depression Ahead. But only a decade earlier, the same author had published The Roaring 2000s: Building The Wealth And Lifestyle You Desire In The Greatest Boom In History.

Indeed, on page 31 of this earlier book, we can still read about "the unprecedented stock boom that almost none of our economists forecast". Better still, "it will continue until about mid-2009, and the Dow will reach at least 21,500 and likely higher."

Sadly, however, while the Dow peaked at 14,093 in October 2007, when we last looked, it had just dropped below 7,000.

In short, forecasts are inevitable and can be useful provided we understand their limitations - and, most importantly, our own. In the midst of a crisis caused by over-optimistic forecasting, we must learn to question our natural tendency to over-pessimism.

In other words, to predict may be human. But believing our own predictions is an entirely different matter. Especially in a crisis.