Lies, damned lies and statistics

[Tizer]

The BBC's `More or Less' radio programme on stats and numbers is back, don't miss it! They take a light-hearted look at the numbers in the news and explode many a myth.
http://www.bbc.co.uk/programmes/b006qshd

[PanBiker]

Panbiker , in terms of number of Asteroids in the normal location ( after Mars ) , most are dust to not much bigger , then there are more sizes but fewer of them , relatively.

Try telling that to the dinosaurs!

[Tizer]

Good one Ian! The dinosaurs were treated to a classic case of the risk being minute but the hazard being (like them) enormous.

[PanBiker]

Aye, I saw a fascinating program a couple of years ago which overlaid impact sites on satellite images of the earth's surface. The big one that did for dino's so to say covers an area about the size of the USA although at the time the impact site itself was where the Pacific ocean is now and just encroaching on the present land mass. The earth is pock marked with numerous impact sites from "smaller" but certainly not insignificant atmosphere beating rocks that made it to the ground. Quite a few of the craters that can be identified by different imagery techniques are considerably larger than the UK. In the vast scale of things, we, the Human Race have only been around for a spit. Fortunately we have not had to experience anything like the impact that wiped out the dinosaurs although the Russians did have the Tunguska Event back in 1908 which certainly would not have been nice if you were taking a walk in the forest that day!

Also in the vast scale of things, it is only a matter of time before another one comes along and contrary to popular sci-fi films and disaster movies, I don't think we could do a lot about it.

[Pluggy]

There was a close 'flyby' of a significant lump of rock recently wasn't there ?

[Stanley]

I go with Hawking who says that the only way we can survive is find another planet to pollute. (I paraphrase) If there was a prize for arrogance and stupidity the human race would be high in the rankings!

[Stanley]

I have been struck by the variation in the reports of the worth of the Face-book share issue. I have heard $1billion, $100million and $99million. One commentator said "the same value as Smith Glaxo Kline". Who is right/ (If anybody is!) Could it be that reporters are firing from the hip?

[Tizer]

Stanley, send your query to the Radio 4 `More or Less' programme and they'll find out for you!

[Stanley]

From later reports I think it may be $104billion..... Not agitated enough to pursue the enquiry Peter.

[Tizer]

There was a very good piece on Radio 4's `More or Less' programme where Tim Hartford analysed the government's claim that a core of about 120,000 families cause much of the disturbance in communities across the UK. The analysis found that the claim was not well founded, not so much on the basis of the numbers but on the assumption that the families identified were causing trouble when in fact they were picked out because they were suffering trouble. Families were selected if they suffered 5 or more out of the following 7 `troubles':
No parent in the family is in work;
Family lives in overcrowded housing;
No parent has any qualifications;
Mother has mental health problems;
At least one parent has a long-standing limiting illness, disability or infirmity;
Family has low income (below 60% of median income);
Family cannot afford a number of food and clothing items.
Thus the survey selected the most `troubled' families in the UK but the government used the results to give the impression that they were the most `troublesome' families...not the same thing.

[Stanley]

Tiz. What a good example of massaging statistics to support an argument. It shows the dangers of using statistics to 'prove' a political point rather and probe for the truth. The insidious thing is that the figures could well be honestly researched and accurate but are wilfully 'misinterpreted'. Is it dishonestly or simple incompetence? Either way, it is misuse.

[Bruff]

Good example here. The other week, some in the media noted that the MP John McConnel had come top of the ballot for Private Members Bills for the 2nd year running and the chances of this happening was several hundreds of thousands, if not millions, to one.

Wrong. This would only be the case if it was predicted in advance that he would top the ballot in each of the two years, and that all MPs would put forward a PMB. Not all did (just 200 or so) and as no one had predicted in advance he would top the ballot two years running the chances were, each time, just 200 or so to one.

It's not just wrong, it's badly wrong with serious possible consequences. You do not multiply these things together, which is what the media did. If you do, and you do it enough times, even quite common outcomes become very rare indeed. A nurse in Holland spent years in jail through just this, as people linked common events (deaths on a ward) to her presence and multiplied up. That is, say a death occurring is 1 in 50, if you're a nurse on duty on 5 of these occasions, 50x50x50x50x50 makes it certain it's down to you, if that's what you want to 'prove'. That lady who was jailed and then released over two cot deaths suffered similarly. I think she killed herself following her acquittal

Richard Broughton

[Tizer]

Same sort of thing with DNA testing for catching criminals. Another example was when the world media made a big issue about two F1 drivers coming in with practice times only milliseconds apart, they thought it must be a chance in a billion. Not so. F1 cars and drivers are so similar, limited by the same factors, all so near the limit, using essentially the same technology and techniques, we should expect them to have very similar practice times.

[Stanley]

Having a statistician as an expert witness for the prosecution sounds frightening to me!

[Tizer]

It's flood time again...and it's time to roll out the ridiculous`once in a 100 years statistic' once more. At least the Environment Agency lady on the radio this morning said they now prefer to use the term `1% risk in a year' which seems more meaningful. But really all these terms are of doubtful value. I'm not quibbling with the value of 1%, 2% etc for risk but with the lack of context - how have they calculated it and to what does it apply? If you apply the statistic to an area of the country it could be meaningless for individual towns; apply it to a town and the same limitation applies to streets and homes. You can have a highly accurate statistic but we need to ask `to what does it apply?'.

[Stanley]

The history of the term 'X year event' goes back to the old River Boards who used the term not as a statistic but as an approximate measure of risk. It was useful shorthand for classifying a risk and prioritising action. In that context was quite sensible. It was never intended to be a firm forecast of risk and it is lazy spokesmen and women who have misused it. I first came across it over 30 years ago when I was dealing with the River Beal at Ellenroad.

[Stanley]

There was a very good example of the limitations of statistics this morning on Today. The government is considering 'improving' the test which determines whether a child is being reared in poverty. The current criterion is if they live in a family that has less than 60% of the National Median income. This of course doesn't measure actual poverty, it is just a handy statistical tool and is relative to other factors. During the discussion about this the statistician giving evidence said that the levels of children in poverty had been falling and was lower now than at any time in the recent past. Great news? Not necessarily. He pointed out that prior to 2010 the decrease could be explained by the New Labour government's benefit policies. The decrease since 2010 is largely due to the fact that the median income has fallen so fewer families qualify, not because of any improvement but because the baseline has shifted. In addition, those families who are in poverty will have been harder hit by the fall in median income because their disposable incomes are spent wholly on necessities. In other words, any politician making statements about child poverty falling on the basis of these figures is being very economical with the truth.

[Stanley]

Have you seen the latest anti-smoking advert on TV? They make the statement "smoking 15 cigarettes will cause a mutation" (of a cell, one of the precursors of cancer). How can they be sure this is true? I could understand it if they qualified it with 'on average'. Bad use of statistics?

[Tizer]

Bad presentation. Many people will think they mean 15 ciggies a day, some will think it refers to the first 15 cigs you smoke in your life, and you're right that it must be based on some sort of calculation of an average. Not very helpful. If information confuses then people will ignore it, and communication fails. A lot of people won't believe it because they think a mutation means you turn into a monster and, after all, Uncle Bill smoked all his life and he didn't end up looking like Shrek!

[Stanley]

I am reminded of Stafford Beers when he told me that it wasn't what was transmitted or how it was received, it was what was understood. The man understood communications.....

[Bruff]

Small number statistics, the next in an occasional series.

The PM is fond of taunting the Opposition when they ask him about food banks, with the line that use of food banks grew 10-fold under Labour. Very true, it sadly did. In 2000, about 3000 folk used them and by 2010, 40,000 did and that is clearly a 10-fold increase.

Now, almost 130,000 folk use them and of course this represents a 3-fold increase since the coalition came to power. We are however, looking at small numbers and so it more sensible (the churlish would say, honest) to note that in the years 2000-10 under Labour use rose by 37K in a decade, but has rocketed and risen by 90K in 2 years under the coalition. 250K are expected to be fed by this route this year.

I thought there was something iffy in the numbers when I heard the PM mention this, but my thanks for the ever reliable Channel 4 FactChecker for the analysis which of course is non-partisan as it quite rightly notes a big increase in the numbers under the previous administration and before the current economic difficulties.

Richard Broughton

[Stanley]

Nice one Richard. So, in percentage terms use of food banks is rising more slowly under the coalition.... A triumph for their economic policies!
I noted another glitch yesterday during the discussions about the proposed HS rail line. Nobody seems to be sure what the initial figure of £35billion refers to, whether it covers the whole project of just the Southern section. I think one thing is fairly certain, it will overshoot the budget!

[Tizer]

An example of Marketing Man's obfuscation this morning on the Today programme. They were talking about sugary foods and the Kellogg's spokesman was responding to the presenter's suggestion that sales of the breakfast cereal had declined. He replied that cereal was still eaten at half of all breakfasts in the UK. This is a crafty way of avoiding having to admit the true story - that sales have dropped dramatically because many people don't eat breakfast now.

[Stanley]

Listening to the apologists for the various food organisations is like a master class in obfuscation. Andrew Opie, spokesman for the food retailers has just been on Farming Today explaining why all is well with imported meat products....
Did you see the enormous consumption figure for fizzy drinks, so big I've forgotten what it was! The Soft Drinks Association say they are doing all they can for the nation's health.......

[Tizer]

A tax on sugary drinks won't prevent people drinking too much, just as a tax on fatty foods doesn't work. The Danes put a tax on `fatty food' but gave it up recently because it had no effect on sales. Imposing a tax simply to make money which is supposed to be then used to counter the effects of sugary drinks is a bit like putting a tax on American hand guns so that you can use the money to bury the victims of gun crime.