Letters : Pooling probabilities
Kidderminster
Your article confuses two distinct uses for statistical techniques (“Faith,
hope and statistics”, 22 November, p 36). One is to estimate the most likely
value for a particular effect, such as the increased survival due to a new drug
treatment, and this is where Bayesian methods seem to offer some interesting
advantages. It is sensible to expect that pooling different kinds of data will
lead to better estimates, even if some contributions are more reliable than
others.
This has nothing to do with probability statistics in the usual sense, whose
results are normally expressed as p-values. Of course it is universally
recognised that the use of threshold values such as 5 per cent for this is
simply a convention.
In the example of the clot-busting drug tested in Scotland, which showed a 49
per cent improvement in a particular trial, the p-value of 0.04 is quoted. This
is not a measure of the reliability of the value of 49 per cent, as suggested,
but an indication that it differs significantly from zero鈥攖he drug has a
positive effect. Other trials would give different values for the average size
of the effect, and the experts’ judgement that the true value could be 20 per
cent might well turn out to be correct, without invalidating the earlier
conclusion.
I find the remark that Bayesian methods may be able to take account of “poor
experimental design” rather worrying. Surely this comes into the sow’s ear/silk
purse area, or perhaps that of rubbish in/rubbish out. No analytical techniques
are going to protect us from either fools or charlatans, and we just have to
hope that in science they are in a minority.
Letters : . . .
Colchester, Essex
Robert Matthews is right to criticise the use of p-values for testing
hypotheses. They date from the time when the most advanced computer available
was the Millionaire, which took 10 seconds to find the product of two
four-figure numbers.
We can now do this more than a million times faster and simple p-values are
not now used by the best statisticians. Instead, they plot the likelihood
ratios, often in graphical form, but they avoid, as far as possible, inserting
the “Bayes factor”.
It is necessary to insert the Bayes factor only when a choice has to be made
between alternative courses of action. Bayesian methods are well suited to
making decisions when these have to be made. But they strictly require us to
assume that we have really considered all the possibilities鈥攖he
probabilities must add up to one.
Likelihood ratios involve only the hypotheses which we are explicitly
comparing, leaving open the question whether there are any hypotheses we haven’t
thought about鈥攁s there nearly always are.
Letters : . . .
Addlestone, Surrey
Of course the idea that science can ever be objective is implausible because
it involves human interaction with whatever is being studied, and as New
杏吧原创 pointed out (“Time’s arrow”, 1 November, p 34): “The act of
making an observation provokes nature into making a choice between contending
realities.”
But scientists, it is often alleged (at least by my wife) live in a world of
their own, thus proving not only the existence of multiple universes but also
that statistics is a purely personal experience.
Letters : . . .
Wokingham, Berkshire
An everyday example of a situation that could be improved by Bayes theorem is
the prosecution of criminals. When a person is on trial the prosecution is not
allowed to present a previous criminal record as evidence. Character witnesses
of “good character” are allowed. The law recognises that a past record of
similar offences must change the likelihood ratio as defined in the article.
The chagrin of a jury who acquits a person only to discover that he has been
found guilty of many similar crimes previously, demonstrates how prior evidence
can change the interpretation of current evidence. Personally, I see a record of
previous convictions as very relevant evidence and would like to see it
allowed.
Letters : . . .
Meersbrook, Sheffield
Drug trials giving false results is perhaps not the most severe problem
created by using p-values.
Social scientists routinely use p-values to assess significance. The results
of their research are used to shape economic, educational and social policy
affecting millions of people. There is both a direct link between academia and
government, and an indirect one when public opinion is informed by research
results (“surveys show that . . .”). If social scientists’ research is
fundamentally flawed, it is likely to be the source of much human
unhappiness.
Letters : A special moment
Tucker, Georgia, US
The gloomy implications of the Copernican argument suggested in “A grim
reckoning” (15 November, p 36, and Letters, 6 December, p 54) need not frighten
us, because we are not at a random point in the life of our species.
It’s unlikely that we would have pondered the implications of the Copernican
argument before we thought of it, and we probably wouldn’t have thought of it
when we were busy spearing mammoths. We couldn’t use it to predict the future of
our species until we had an idea of the age of our species, which we are only
starting to confidently estimate. And in a thousand years the idea will be old
news and nobody will excitedly propose its implications in an article in New
杏吧原创. We should hardly then be surprised that we are discussing these
implications now, after the development of classical philosophy but before we
can engage in interstellar travel.
Believing that we are at a random point in our development, like believing we
are governed by a set of randomly chosen physical laws, is not humility, it is
conceit.
Letters : Some way to go
Thank-you for the report of my talk at the recent Royal Society discussion
meeting on quantum computation (This Week, 15 November, p 10). Unfortunately, I
was either unclear or misunderstood by your reporter. I am quoted as saying that
Teramac is a hundred times faster than any machine currently available and
provides computing power equivalent to 100 billion Pentium processors. The
article goes on to say that Teramac is made up of a network of quantum dots.
These statements are not correct.
Teramac is an experimental computer built at Hewlett-Packard Laboratories
using standard silicon technology. It is based on Field Programmable Gate
Arrays, which were the subject of the article on page 30 of the same issue of
New 杏吧原创 (“Creatures from primordial silicon”). Teramac is 100
times faster than any currently available workstation computer for solving
certain problems (but not any currently available machine).
The unique aspect of Teramac is that it can operate very efficiently even
though it contains a large number (about 220 000) defects, any one of which
would make a conventional computer inoperable. Thus, we view the architecture
upon which Teramac is based as a good model for future molecularly assembled
computers, which will have components (quantum dots) as small as a few
nanometres across and will necessarily contain a high fraction of fabrication
mistakes.
These molecularly assembled computers are theoretically capable of
approaching the physical limits of irreversible computation, which correspond to
the computational speed of 100 billion Pentium processors with the expenditure
of only one watt of electrical power.
Although I did not state this explicitly in my talk, I do not expect
computers with this computational capacity to be available until the middle of
the next century.
Letters : Tiananmen taint
Leeds
Andreas Frew’s remarks about Jiang Zemin in Washington Diary (29 November, p
58) are misplaced. Jiang was in the leadership of Shanghai until after the
protest movement had been repressed in Tiananmen Square on 4 June 1989. He was
not even remotely responsible for the decision to use force against the
protesters in Beijing. The way he dealt with a similar movement in Shanghai has
been hailed as peaceful. His admission of “mistakes” on the part of the Party
with regard to the repression was significant, not a “platitude”.
Sniping remarks about Jiang Zemin obscure the real problems in US-China
relations, a field in the US in dire need of more knowledge and less
rhetoric.
Letters : Juicy analogy
Rochdale, Lancashire
Discerning the meaning of Sainsbury’s “Now even juicier” claim for its orange
juice can perhaps be approached by analogy (Feedback, 22 November).
If “now even juicier” were applied to an orange it would clearly mean that it tended
to drip more when you ate it鈥攊n other words, it had a higher water content than
some notional standard orange of the similar diameter or weight.
So is Sainsbury’s putting more than 1 liter into the carton, or more water into the
litre? Or does the claim mean something else? I think we should be told.
Letters : Criminals and chemists
Excellent@btinternet.com
Feedback is right. I was cut to the quick by the slogan “Chemistry: nobody
will steal your books”, proposed as a slogan to attract more chemistry students
into university (29 November). I’m a female chemist working on an Open
University preparatory course and was extremely upset when the bag with my
textbooks, notebooks and nearly completed assignment was stolen from my car a
month ago.
If any New 杏吧原创 reader in the Slough/Maidenhead area has found
these books, please hand them in to the police so I can finish my work.
Meanwhile, I think a great way to get more chemistry students is to use the
slogan, “Chemistry: even thieves will want your books”.
Letters : Justice for apes
Reading, Berkshire
In a compelling article, John Blatchford suggests that nonhuman great apes be
considered as people (Forum, 29 November, p 56). This would imply that they
could be deemed owners of their own property鈥攖he areas in which they
currently subsist.
Blatchford is not alone in challenging the existing moral order. The Great
Ape Project has been promoting the personhood of great apes since it was founded
in 1993. This campaign has arisen from a paradigm shift over the past 20 years
or so in our understanding of the complex emotional and mental lives of the
great apes鈥攁 complexity that demands we confer the basic rights of life,
individual liberty and freedom from torture on all great apes.
The Great Ape Project is campaigning to make these fundamental protections
part of national and international law. As Blatchford points out, it is
envisaged that legal actions would be mounted by advocates, as nonhumans are not
able to do this themselves鈥攖his is analogous to the way that human
children’s rights are represented by legal advocates in the courts.
Blatchford’s ideas may seem radical, but to an increasing number of
scientists, philosophers and laypersons, the logic of the moral arguments that
he articulates seems inescapable. Indeed, such is the force of the ethical
issues raised by investigations into great apes’ cognition that Britain’s Home
Office has recently banned scientific experiments on great apes. This
little-noticed part of the statement on 6 November banning cosmetic tests on
animals represents a profound alteration in our attitudes to other sentient
beings, and a significant step forward in recognising the rights of great apes
as people.
The Great Ape Project’s Web site is at www.envirolink.org/orgs/
gap/gaphome.html
Letters : Petrol and pollution
Arlington, Virginia
Cleaner-burning petrol containing methyl tertiary-butyl ether (MTBE)
is providing California with the best air quality in four decades (“The dirt on
clean fuel”, 22 November, p 24). In Los Angeles, for example, average ozone
levels in the summer of 1996 were 18 per cent lower than in 1994 and 1995, after
adjusting for weather differences. In 1996, this region had a record low of
seven smog alerts for the entire year. By way of comparison, in 1970 Los Angeles
had no fewer than 148 first-stage smog alerts.
Your article says that the US Environmental Protection Agency has classified
MTBE as a potential carcinogen. But a European task force on MTBE, convened by
the European Centre for Ecotoxicology and Toxicology of Chemicals, concluded
this summer that, according to the criteria set forth in the European Union’s
Directive on Dangerous Substances, MTBE is not carcinogenic.
The article quotes a New Jersey researcher as suggesting MTBE can “probably
penetrate people’s skin during washing. It could then accumulate in fatty
tissues”. In fact, pharmacokinetic tests demonstrated that MTBE is poorly
absorbed through skin and dermal exposure would not pose any increased health
risks.
Bergeron suggests that both Europe and the US face a Hobson’s choice
regarding cleaner burning fuels鈥攅ither clean water or clean air. “The
quest for both remains as elusive as ever,” he opines at the end of the
article.
The reality is that the US is enjoying the best air quality it has seen in
decades, and at a time when auto use is at an all-time high. Tough federal
regulations which require the upgrading, replacement and monitoring of all
underground storage tanks by next December will go a long way to ensure that
gasoline, with all of its hundreds of compounds鈥攎any of which are far more
toxic than MTBE鈥攚ill stay out of the groundwater.
Bergeron correctly notes that most European tanks are double-walled, so
Europe is already ahead of the game. Both the US and Europe need not sacrifice
air quality for water quality. Both are possible with cleaner-burning fuel
programmes.
Letters : . . .
Alsager, Staffordshire
If ethanol really is a “known carcinogen and reproductive toxin” in the US,
then a speedy return to prohibition would seem to be indicated.
Letters : . . .
Lou Bergeron writes: According to California’s Safe Drinking Water
and Toxic Enforcement Act of 1986, ethyl alcohol (ethanol) in alcoholic
beverages is known to cause reproductive toxicity, and alcohol abuse is known to
cause cancer.
James W. Embree, a toxicologist with Geomatrix Consultants in San Francisco,
confirms that ethanol is known to cause liver cancer when ingested at high
levels. Embree, who has been involved with the toxicology of MTBE for 20 years,
also says, “If you held the same magnifying glass to drinking alcohol that’s
being held on MTBE, ethanol comes out the loser.”
However, he also says that from a toxicological point of view, “At the low
exposure levels you would see when these compounds are used in gasoline, neither
one’s a problem.”
Letters : Glory of gloop
London
Andy Robinson suggests that the hiss of an amplifier would be as good a
source of randomness as lava lamps are (Letters, 29 November, p 63). The problem
is that an amplifier is rather prone to external influences (as is obvious when
you hear your local taxi firm over the top of your favourite CD), so the
randomness it produces is of dubious quality. An attacker need only place a
sufficiently powerful broadband transmitter in the vicinity of your “random
source” and suddenly they will be in complete control of your “random” numbers.
They could even transmit stuff that sounds like white noise, but is in fact
completely known to them鈥攕o it isn’t even possible to detect their
influence.
The beauty of a lava lamp as a source of randomness is that it is difficult
to imagine how an outsider could influence the pattern of rising and falling
gloop. Unfortunately, the same cannot be said of the TV camera (and associated
amplifiers) pointed at it.
Of course, sources of quantum noise, such as noisy diodes, are commonly used
as a source of randomness, but they must be carefully designed to avoid outside
influence.
Letters : Paper power
Aylesford, Kent
Your article was unfair on recycling and Aylesford Newsprint (“Burn me”, 22
November, p 30, and Letters, 13 December, p 55).
The sludge produced at Aylesford Newsprint contains very small concentrations
of heavy metals and is not required by the Environment Agency to be registered
as a special waste鈥攚hich, in turn, would have created a requirement for it
to be disposed of at specially designated landfill sites.
The energy created by burning the sludge generates 20 per cent of the Mill’s
steam requirements and about 3 megawatts of electricity. This reduces the energy
requirement from fossil fuels as supplied by the National Grid.
In conjunction with National Power, Aylesford Newsprint operate a Power and
Steam facility supplying the Newsprint business and additionally, two large
neighbouring manufacturers. The energy figures quoted in your article do not
specifically relate to newsprint.
The comment regarding the 4.19 million kilometres in collecting used paper is
correct but overlooks the average of six miles per tonne and the comparison of
importing from Canada or Scandinavia, the alternative sources for suitable
fibre.
In addition, it fails to identify the fact that there are currently only
seven municipal waste incinerators in the country, some of which have had to
undergo rebuilding to enable them to continue operation under European Union
regulations. The contention that an incinerator is a better option is a tenuous
one with the present situation of planning and public awareness.
Aylesford Newsprint does not dispute the value of incineration with energy
recovery in the disposal of exhausted and contaminated paper materials but
believes that it is preferable to recycle where practical, achieving the benefit
of both activities.
Aylesford Newsprint has provided the public with information about the
environmental aspects of its operations over the last three years through the
production of its environmental report. In these the company has provided the
audience with detailed information on the way it manages environmental issues
and has identified where it has improved its approach through best available
technology and provided a state-of-the-art operation.