Portal de Periódicos da CAPES

Letters/Editorial

2009; Oxford University Press; Volume: 6; Issue: 3 Linguagem: Inglês

10.1111/j.1740-9713.2009.00382.x

1740-9713

Demographic Trends and Gender Preferences

SignificanceVolume 6, Issue 3 p. 140-141 Letters/EditorialFree Access Letters/Editorial First published: 24 August 2009 https://doi.org/10.1111/j.1740-9713.2009.00382.xAboutSectionsPDF ToolsRequest permissionExport citationAdd to favoritesTrack citation ShareShare Give accessShare full text accessShare full-text accessPlease review our Terms and Conditions of Use and check box below to share full-text version of article.I have read and accept the Wiley Online Library Terms and Conditions of UseShareable LinkUse the link below to share a full-text version of this article with your friends and colleagues. Learn more.Copy URL Share a linkShare onFacebookTwitterLinked InRedditWechat Statisticians owe a lot to Darwin. Apart from anything else we are human beings and therefore products of evolution; knowing where we have come from is always a bonus. We owe him also, not all that indirectly, our discipline. Karl Pearson, father of modern statistics, sired that progeny through being asked for help in analysing, of all things, the sizes of shore crabs in the Bay of Naples. The man who asked him, W. F. R. Weldon, was an early follower of Darwin and wanted to establish whether the crabs were evolving into two species. The request led to Pearson shifting his professional interests from physics to biometrics; he mathematicised the theory of evolution. Modern mathematical statistics resulted. Moving considerably down the scale in the great scheme of things, we owe Darwin a large number of our jobs. Genomics, molecular biology, biotechnology, pharmacology and agricultural science are fields that depend absolutely on statistical analysis to get anywhere at all; all of them depend for their core theory upon his insights without which some would not exist at all. Not that the role of statisticians in the development of evolutionary theory has been entirely beneficial. Francis Galton, the Victorian statistician who gave us correlation and regression to the mean, also gave us, with the most high-minded of aims, eugenics—the plan to improve the human race by encouraging the fittest to breed and by “dissuading” the “unfit” from doing so. We impose our own interpretation on “the fittest”—it does not mean “people like us”. “Fittest” does not mean most cultured, most intelligent, most intellectual, most beautiful or most complex. It means those who spread their genes most widely. Former Prime Minister Arthur Balfour spotted this early on, pointing out to the First International Congress of Eugenics in 1912 that fairly fundamental flaw in their reasoning. H. G. Wells, in The Time Machine, took the same point to its logical conclusion: the future “successful” descendants of the current human race were not to be the gentle, fruit-eating race of sun-lovers that his time-traveller met and loved, but the cannibalistic, underground, daylight-shunning Morlocks who fed off them—descendents, as he imagined, of the light-starved, culture-starved Victorian working class. In fiction, as in fact, the noble aims of the eugenicists descended into horrors. The greatest ideas are the ones that are simple but which reveal complexities of insight the deeper you delve into them. Thus it is with Darwin. “Survival of the fittest” is, at first sight, so simple that modern philosophers of science have questioned whether it is anything more than a tautology, whether it deserves the name of scientific hypothesis at all. At second sight it offers complexities that 150 years later still have specialists disagreeing passionately among themselves. Experts such as S. J. Gould and Richard Dawkins interpreted Darwin for general readers—but disagreed over whether evolution proceeded via “punctuated equilibrium” or selfish genes. Non-specialists do not have to go down the creative madnesses of the Creationists to find difficulties with “survival of the fittest”. It certainly does not mean a necessary rise to ever-more-complex, ever-more intelligent, ever “higher” life forms. There was nothing inevitable about the emergence of humanoid life, or of intelligent life (leaving aside the question of whether the two are synonymous); there is nothing inevitable about their continued survival. Statistics is extracting what data means from a large amount of raw material that can be overwhelming in its volume and undigested state. Nature fits that description. Darwin looked at nature's data—shapes of finches’ beaks, of tortoises’ shells, at fancy pigeon varieties, at fossils and coral reefs and worms—that had been there all along for others to see and he drew his conclusions. All of nature is still out there to be looked at—for now. What we see—what phenology and population statistics of endangered species and analyses of climate change and ice-sheet melts and declines in biodiversity all tell us—is that a planet's ecosystem is in crisis. The statisticians’ job now is to monitor that crisis, to forecast likely outcomes and to analyse the steps that can be taken to mitigate it. Life on earth is wonderfully diverse. It would be a shame if only fossils were left to tell of it. Julian Champkin Private school “sacrifice”? It would be a little surprising to find that for their “sacrifice” for paying school fees parents were not ensuring some advantage in their offspring jumping a few places up the queue of later career ladders (Significance, June 2009). What readers may not know, however, is that the very same source that Francis Green and colleagues used to derive this finding can also be used to find that there are potential disadvantages in sending your children to an independent school other than narrowing their social horizons and damaging your bank balance. A study colleagues and I undertook using the British Household Panel Study (BHPS) a decade ago resulted in findings that suggested that if you set (and sent) your offspring up to succeed, and they did not, they may be more likely to blame themselves. Men in the BHPS who attended private school and who were born in affluent areas moved to poorer areas more often than did children with other education from low mortality to high mortality areas, joining a group of male migrants with poorer levels of mental health and higher later premature mortality1. Sample size was so small then that the confidence limits were wide, but hopefully someone else will repeat and extend the study now more years have passed. Should you be in the very small group able to afford private education and be thinking of taking that risk for your children, it may be better to be aware that they may not thank you should all not go well later on. Should you no longer be able to make that decision due to the current economic situation you can at least console yourself that you may not have been making the right decision anyway, depending on the extent, as Clint Eastwood once said, to which “you are feeling lucky”. Incidentally, does anyone know why paying for private health care to jump your place for treatment in the queue is hardly ever referred to as a “sacrifice”, but the same behaviour in education is? Danny Dorling Sheffield Dr Fisher's p-values In Significance (June 2009), Dr Fisher appears to have had a change of heart over the usefulness of the p-value. In particular, he (apologies if “he” is a she) implies that in the case of a cross-over trial p-values are of more use than the estimate, even, he implies, when this is an interval estimate. His point is that since this study may not give an effect size that is predictive of that likely to occur in long term use of the drug, plus the potential for carryover we should not be concerned with the size of the effect, just whether an effect has occurred at all. Although this argument has some intuitive appeal, I believe it not only can be shown to be flawed under closer scrutiny, but also discourages the critically important scientific discussion that must occur around the effect size of interest. If we accept the rationale given for focusing on the p-value, then I do not see why this argument is specific to the cross-over trial. For example all pre-clinical studies are much less predictive of the future use of the drug, irrespective of the design used, and there are other sources of error than carryover that are just as likely to lead to bias. The same situation exists in virtually all other areas of product development, not just in the pharmaceutical industry. Hence, his argument, I suggest, implies that we should be more often than not primarily focusing on the p-value, which is inconsistent with what Dr Fisher explicitly states. The problem with his argument is that, although we may well have only a weak understanding of how the results of the current trial predict future drug performance, it is not the case that any effect size, however small, will be considered sufficient to progress to the next stage of development. This becomes clear if we imagine running the perfect study resulting in knowing the true effect exactly; outside of the bluest of blue sky laboratories there will be an effect size so small, but non-zero, that will not justify the additional costs of taking the drug further. We do not even have to do this thought experiment. When the study is designed the project team are forced to think about what size of effect they are prepared to “miss” when powering the study; even if such formal power calculations are not done, there is an implied “missable” effect size. In many cases, particularly in clinical trials where studies are very carefully powered, it is often sufficient to simply require superiority to the negative control as the ultimate decision criterion, which can be crudely interpreted as focusing on a p-value. But this is not because the effect size is irrelevant, but precisely because the effect size was carefully considered when powering the study. More generally, it is very important that the project team consider whether the decision criteria for “study success” be based on some confidence that the effect is not only bigger than zero, but also of some meaningful magnitude (this latter threshold often being set by reference to a positive control). I do not see the fact that we are running a cross-over trial as opposed to any other study design is relevant to these arguments. So, given my criticism of Dr Fisher's defence of the p-value, it leads me to wonder if there is a place for this resilient study summary. I think there is—and this place is called a museum. Phil Woodward Kent Error of our ways Where do residuals go to drink? In Error bars. Letters should be sent by e-mail to significance@rss.org.uk, or by post to: Significance Letters Page, Royal Statistical Society, 12 Errol Street, London, EC1Y 8LX. They should be short (preferably under 250 words), may be edited for length and should clearly indicate whether or not they are for publication. They must be received by October 15th, 2009, in order to be considered for publication in the December issue. Michael Wallace Reference 1Brimblecombe, N., Dorling, D. and Shaw, M. (2000) Migration and geographical inequalities in health in Britain: an exploration of the lifetime socio-economic characteristics of migrants. Social Science and Medicine, 50, 861– 878. CrossrefCASPubMedWeb of Science®Google Scholar Volume6, Issue3Special Issue: Darwin 200th AnniversarySeptember 2009Pages 140-141 ReferencesRelatedInformation