Tuesday, 16 November 2010

Statistics, Correlation – And Stupidity

Government at workIT IS SAID that, in the middle of the 20th Century, both the Russians and the Americans conducted a programme of psychic investigations in order to establish if such ‘powers’ could be harnessed in their espionage activities. And it is also reported that both sides, despite reaching no firm conclusions in the matter, nonetheless achieved some surprising results.

One of the most interesting findings was from establishing a ‘coven;’ consisting of 13 individuals whose leader would set its members a task – such as accurately predicting the number of individuals recorded in a selected telephone book, or how many cookies there were in a jar.

What they found was that the 12 guesses (formulated through deep meditation, trance, or simple concentration) differed widely from one individual to another. But, if the results were averaged (by totalling the results and dividing by twelve) then, on many occasions, the calculation proved to be absolutely right.

They also found that these accurate results were more frequent than could be rationally explained by chance.

Perhaps Government should return to its ancient, pagan roots and establish a number of specialised covens to develop its policies - in preference to the modern method of relying upon computer modelling and stupid statisticians…

Over the last twenty years, I have found that researching any article has become harder and harder. Politicians evade simple questions; business leaders now direct journalists to their PR departments rather than handle enquiries themselves; and the majority of the general public are unwilling to go on-the-record for a controversial piece. On the other hand, it seems that everyone is prepared to release carefully contrived statements to attract media attention – rather than purchasing a classified ad.

When I first established the Canvey Beat, it took just a few days before my inbox was inundated with press releases from various interest and pressure groups determined that I should accept all their statements as facts – and publish their contents in an article. It seems that the How to get the media to work for you free PDF downloads have never been more numerous or hungrily read.

Modern statistics, it seems, has now firmly replaced the ancient tradition of casting runes or studying a freshly killed rabbit’s entrails. Virtually all interest groups now refer to ‘facts’ that are apparently derived from numerical figures that can be implicitly trusted because they have been computer generated. Indeed, it appears that, without a set of figures to support one’s argument: no one is about to take you seriously. As a result, many journalists find they spend more of their precious time trying to get behind a set of figures than they do interviewing parties concerned. And the one truth that has emerged from my personal investigations over the years is that: ALL FIGURES LIE.

Last week I was so disgusted with Dave Blackwell’s attempts to manage an Overview and Scrutiny Committee meeting at Castle Point Borough Council that I took refuge in choosing something else to write about. I decided that anything I wrote on the meeting would be so acerbic that it would be bound to result in a libel action that I was in no mood to participate in. Fortunately, the Echo too chose to ignore the matter, and I was not obliged to point-out their inevitable bias in a follow-up piece.

I calmed myself by writing, here, about the UK’s £4,8 trillion of debt, reported by my favourite TV broadcaster, Channel 4; and followed-up a local Canvey Beat piece with another on the same subject (but from a national angle) on my new British Chronicle Blog.

It was in developing the latter piece, regarding Rebecca Harris and her Private Member’s Bill, that I was drawn to the obvious inconsistency in the statistics being banded around.

Fatal road accidents since 1926I had taken the trouble to investigate the number of historical road deaths for myself – and the most reliable source I could uncover was the Government’s own statistics, in a briefing paper, updated on 25th January this year.

Those figures showed a high correlation between artificially extending the hours of daylight in the evening – and road accident deaths. It is so strikingly obvious (that road deaths increase whenever the UK adjusts its time zone) that I was unable to reconcile those raw, accurate figures with the claims being made by the Lighter Later campaign group, and the Royal Society for the Prevention of Accidents (RoSPA), who stated that putting the clocks forward another hour would actually save lives.

It was an anomaly that I was determined to pursue, so I contacted my MP, who just happens to be Rebecca Harris, and she (unlike her predecessor who always reverted to including in his replies: ‘The following is not for publication and is strictly private and confidential’ – and then inevitably went on to threaten legal action over any unfair criticism that might be published) immediately responded with a friendly note with no OTRs – and then followed-up with all the supporting evidence she had gathered from her parliamentary computer.

Such transparency is exceedingly rare.

I must admit that I intentionally ignored a lot of her information regarding environmental claims – such as potential energy saving. Instead, I focused upon the RoSPA report for which I could find no explanation.

There is a common adage that there are lies, damn lies – and statistics. And what I found confirmed that fact.

RoSPA does not lie in its conclusions that, when the clocks are moved forward one hour, more lives are apparently saved in the evening than the small increase in morning casualties. But it completely ignores the strong correlation between introducing more daylight saving and TOTAL road deaths.

We all know that alcohol is responsible for many road related casualties – and that more alcohol is consumed during the summer months when evenings are lighter; but that evidence, easily gleaned from government statistics, is not once mentioned in RoSPA’s report.

RoSPA’s claims are based upon a 1998 report, commissioned by the then Government, whose authors are not mentioned. Accordingly, it seems to have been produced by some Parliamentary committee which was at pains to produce a conclusion recommending the adoption of Single Double Summer Time despite the contrary evidence that was immediately evident in the raw statistics. And the way that this has been done is the very way in which others ‘twist’ numerical data to suit there own political arguments.

The previous Labour government was always keen to promote good economic news – by always ensuring that the statistical set, used to produce their misleading figures, only spanned the best numbers from which only a good result could be obtained. In the same way, the report’s figures had been engineered to only compare those instances which would prove the SDST case.

The first thing we need to recognise is that road deaths are not predictable. By their very nature they are strictly random events – particularly when it comes to the actual time of day in which they occur. Road accidents have much more to do with weather conditions; driver fatigue and driver intoxication than they do with whether they occur during the morning or evening rush hours. The only part that light plays, it could be argued, is during the periods of sunrise and sunset when, depending upon the way in which you are travelling, the sun may be in your eyes. And that condition is endured every day - no matter what time zone is employed.

The 1998 report extracts a small subset of individual road deaths from the daily totals by only using those occurring between 7.00am and 10.00am – and 4.00pm and 7.00pm. The fact that there are far more deaths incurred between 7.00pm and 7.00am each day is totally ignored.

The reports anonymous authors, and RoSPA, conclude that SDST should be supported; because this contrived evidence makes it clear that, during additional daylight saving time, more lives are saved between 4.00pm and 7.00pm than are additionally suffered in the morning period of 7.00am to 10.00am.

The report only analyses two three-hour periods, six hours in total, out of a day’s length of 24. That is, it only uses a quarter of the evidence available to it. And that very 75 per cent of data, which has been excluded, ‘proves’ the opposite case. (Just as using one-hundred per cent of the available data does).

Single Double Summer TimeBut, in support of the theory that the 1998 figures have been artificially manufactured: here is another curious aspect of that report’s data set.

Inset is a graphic I developed for my British Chronicle piece, and I reproduce it here so that readers can examine the times of sunrise and sunset. (There is a small difference in these times each year – but only by something like a minute).

Here we need just to consider the GMT+1 table – upon which Britain’s clocks were operating in the 1968-1971 trial and upon which the 1998 report is based.

The 1998 report states that the morning data was based upon the period 7.00am to 10.00am; but, as can be clearly seen from the table, for six months (April through September) during each of those year’s data it was already light before 7.00am arrived. (And do not forget that there was no putting forward; or putting back of clocks during this period to confuse the statisticians – because Britain remained on British Summer Time (BST) throughout the trial).

As for that evening rush hour period, from 16:00hrs to 19:00hrs, the same error appears. The selected period does not correctly straddle the hours in which sunset occurred. For half the data selected, pitch darkness had already arrived.

Data selection appears to be so arbitrary that it is meaningless.

For the study to have made any sense at all: it should have confined its data set to the road deaths occurring during the extra hour of darkness in the morning (just before dawn) and the extra hour of light provided in the evening (immediately preceding sunset).

That is what the public currently assume to be the case when RoSPA states, categorically, that more lives will be saved in the extra hour of daylight than will be additionally incurred by the extra hour of morning darkness. But neither situation has ever been examined.

That extra light hour does not take place during 4.00pm and 7.00pm; but during the last hour before sunset – whose earliest time is 3.52pm and whose latest is 8.18pm (in London). And that extra dark hour does not take place between 7.00am and 10.00am; but between the hours of 3.50am and 8.09am in the morning (in London).

And those qualifications, in parentheses, mark the pure stupidity in all the figures used by the 1998 report. You cannot confine yourself to a single time period, across the whole of the country, to base your conclusions upon; because the sun sets, and rises, at different times depending upon the location.

The figures can only be taken from a data set that, in the information recorded for each individual road death, indicates that the accident occurred either ‘in complete darkness’ or ‘in complete light.’ And only from a data set that confines itself to those examples within an hour of sunset or sunrise at the actual accident’s location.

Anything less and the whole exercise is meaningless.

So, using the proven facts detailed above (that half the data set is completely erroneous) approximately half of those apparent dark morning deaths alluded to in the report actually took place in full daylight. And some fifty-per cent of the ‘saved lives’ due to the lighter evenings, actually took place when it was pitch black. If we were to remove all those inconsistencies from the report’s findings: is it the case that there was, in fact, no statistical difference in deaths between the morning and evening periods? And can we therefore conclude that changing the clocks has no effect whatsoever on darker morning or lighter evening road accidents; and that all additional deaths are simply due to increased daylight, after 7.00pm, in which we British can enjoy an extra pint?

Responsible statisticians will be appalled at the methods employed in the 1998 report. An obvious correlation appears to have been conveniently ignored for political ends and the data set does not reflect the light conditions being examined. But just what purpose has it served? - other than to assist Labour’s friends in the leisure industry to further improve its already healthy profits from that Government’s calamitous relaxation of licensing hours.

When Rebecca’s Bill comes before the House on 3rd December, I personally hope that it will be passed. Not because I am in favour of a move to SDST (I am definitely not) but I do want this new Parliament to thoroughly investigate the claims that have been made by the previous Government, and RoSPA, to discover if this is just a case of common incompetence – or if there is more to this matter than meets the eye.

In the meantime, perhaps those basing their arguments on similar unfounded statistics would be better advised to form their own covens upon which to base their conclusions.

And if Peter Hitchens, or the Mail on Sunday, are reading this: they may wish to redeem themselves and their British Time campaign by taking up the evidence presented here and conducting some worthwhile journalism…

… (28/11/2010, Mail Online) – Portugal warns Britain: We switched to Berlin time and it was a catastrophe

No comments:

Post a Comment