The Social Research Hub and I recently had a Twitter exchange about whether sociology and social research PhD studentships should be accompanied by some kind of career planning. Since that day, I have read yet more despairing tweets about the fate of people who have completed PhDs in sociology and do not find a place to fit into the workforce.
So I said that I would write a bit about the research training that we have been doing in the ESRC International Centre for Life Course Studies in Health http://www.ucl.ac.uk/icls/training.
Our wish to set up a Centre was strongly motivated by the very opposite problem to that of which twitter-writers have been complaining. Over at least 10 years, members of our research group had been struggling to find people to work with us on projects investigating health, wellbeing and resilience over the life course. The possibilities to do this are increasing at an exponential rate due to the maturing of Britain's birth cohort studies and other longitudinal studies such as the British Household Panel Study, and the English Longitudinal Study of Ageing. More recently, large amounts of funding have been awarded for a new, enormous panel study similar to BHPS but 4 times the size, and for a brand new birth cohort study which will also produce a very large amount of data. Anyone who wants to find out how to acquire these data has only to go to http://data-archive.ac.uk/find.
So here are all these data, but where are the people trained in sociology who can use them? For some reason, the tradition of data analysis was lost from sociology between the time when I studied for my BA and the time I returned to academic work in medical sociology. So much so that many people would not regard me as a medical sociologist at all but as a 'social epidemiologist'. But I never believed that to understand the social determinants of health and wellbeing you needed to be an epidemiologist. What you needed was a research group where people knew their stuff in (a) sociology (b) social history (c) human biology (d) developmental psychology and (e) statistics. What you are doing is puting togther the ways in which people develop and change over their lives in terms that can only be understood with a knowledge of (a) to (d), formulated into ideas that will be tested using (e).
ICLS training does not aim to make people bigger experts in the separate disciplines than they already need to be to get a good Masters degree. But learning how to pull together evidence from different spheres and use it to formulate and test the relatively complex hypotheses you need in lifecourse studies is a very transferable skill! In way it is quite the opposite to some of the very arcane, specific topics often followed in PhDs, both in the sciences and the social sciences. Because many of our topics are policy relevant ("Does the divorce of your parents make a difference to your mental health once you are adult?" "Is it bad for children if their mothers have paid employment?" "Does the kind of job you worked in make a difference to how healthy and independent you can remain in your older years?""What is the effect on your mental wellbeing of losing your job?") ICLS PhD candidates also get thrown in the deep end in this respect. One of the many great things about them, which also reflects some other Twitter-exchanges I have been having recently, is that they are natural communicators. By the time I was getting told by my funders that I needed to do 'public engagement' I had become far too much of a nerd. But our students just took to relating to non-academic partners in our enterprise like ducks to water. What THIS in turn means is that they begin their careers being (a) in demand for their knowledge of what to do with a complex data set (b) used to explaining their results in clear language (c) knowing some people outside academe who will be likely to keep them in close touch with what non-academics actually want to know about.
Now, if this sounds like a load of off-putting statistical stuff, all I can say is that the first crop of ICLS PhDs came with a wide range of statistical understanding, many with very little indeed. I myself never even took the Maths "O" level (that is how old I am) and got into data analysis due to the enthusiasm and persistence of colleagues over my working life. Learning what to do with a data set is not a matter of mathematics at all. It is a bit like learning to drive a car (I never learned that either).
But what IS a problem is that there do not seem to be many sociology degrees that might be relevant to the training we give. Many sociology degrees do not seem to deal with, for example, social structure, family change, social institutions. It is hard to find anyone teaching basic Weber, Durkheim, Marx, Bourdieu, etc. Or, more up to date topics relevant to contemporary social change, but from a sociological perspective looking at institutions, identity etc. I am sure that many sociologists looking at what emerges from the large longitudinal data sets at present would be able to think of other questions, other angles. That is badly needed.
Saturday, 1 June 2013
Wednesday, 17 April 2013
Death of the Phillips curve?
As unemployment fell, economic inactivity rose steadily. You can see that the opposite trends expected between inflation and unemployment are not there. But the opposing trend with inflation is there if you look at economic inactivity.
The paper in which this grph appeared first was considered outrageous by social policy and sociology journals and never actually got published. I am too lazy to update it now!
Friday, 12 April 2013
Unemployment statistics: not the whole story
There have been many very interesting posts of economic trend data following the death of Mrs Thatcher. one statistic I have seen less often is the economic inactivity rate. For those who don't think a whole lot about employment figures, being 'economically inactive' is not the same as unemployed. In fact what is officially called 'the economically active' population consists of the employed, the self employed and the unemployed. You are 'active' in the labour market as long as you are looking for work,even if you didn't find any. The corollary of this is that the unemployment rate is not the % of the whole working age population who are unemployed, it is the % of the economically active population. Then there are the economically inactive, who are all those who are not looking for work. The most common reason to be inactive is that you are looking after the home and family, and this (still) mostly applies to women. The second most common reason is what used to be called 'permanent sickness', i.e. long term ill health bad enough to mean a person is not able to work. This 'permanent sickness' is what all the new rules about claiming out-of-work benefits is about. The third most common reason is early retirement. Over time, what has happened is unemployment fell and stayed low, while permanent sickness and early retirement rose. By 2005 or thereabouts, an unemployment rate of around 6% co-existed with an economic inactivity rate of 16% in men of working age.
Finally got the diagram into this document, phew! What a performance. What I hope it shows is how after 1995 or so, unemployment fell dramatically while economic inactivity continued to rise. This diagram is for men only, so this has nothing to do with domestic duties. Nor has it got anything to do with serious ill health, as at this same time life expectancy in men was rising fast.
Thursday, 4 April 2013
A new measure of "social class"
There you go! In the first paragraph of Mike Savage et al's paper on a new model of social class they cite Richard Wilkinson and Kate Pickett as authorities on health inequality. As pointed out in my last blog, Wilkinson's work has never been about health inequality.
Anyone who read my last blog will also anticipate my reaction to using the term 'social class' to refer to cultural dimensions of social inequality. I just think it is so much less confusing to stick to calling social class what the official UK Statistics Office has defined and validated, i.e. employment relations and conditions.
But OK, let's call the new measure a measure of social position, the general term I like to use to refer to all dimensions of inequality whether they be cultural, economic or occupational.
It is great that the authors start out from a sound knowledge of Goldthorpe and colleagues clear definition and measurement of social class which has provided the present official UK measure (the NS-SEC). Why after all this do they think we need another measure?
Firstly, the NS-SEC does badly in predicting cultural activities and identities. The simple answer to this is that it is not supposed to. Whether or not a class measure based on occupation is related to cultural identity is an empirical matter. Secondly, the NS-SEC does not tell us who is in the 'elite'. I think anyone who uses it would agree with this. Thirdly, economists don't like the NS-SEC. I can surely attest to the truth of this: no record of occupation was recorded in the whole of the English Longitudinal Study of Aging (ELSA) life-grid while the sociologists took their eye off the design. Economists only think of income and wealth, not social class, which is fine, that is their discipline. Another criticism of the Erikson-Goldthorpe-Portocarrero (EGP) class schema (from which the NS-SEC is derived but in a rather distant form) is that it overdoes the manual-non-manual divide. I think John Goldthorpe would agree with this, I certainly do, and it is one reason why the NS-SEC does not have a manual-non-manual divide at all.
Perhaps the ultimate irony of the study is that the authors actually used the NS-SEC to validate their sample distribution! (page 6 Table 1). And this is the correct schema too, not the old EGP, and so it does not have any 'manual workers', or indeed 'skilled or 'unskilled' workers'. Managers are divided into 3 classes and professionals into 2. On the basis that the BBC audience survey (161,000 people) was biased (compared to what I wonder?) the BBC did a face to face survey with a sample of 1026 chosen to be nationally representative. It was on this representative sample that the authors carried out their multiple classification analysis of cultural capital rather than the much bigger BBC audience sample.
Seven classes emerge from the latent class analysis, the result of allowing income, wealth, house price, number and average social prestige of social contacts, highbrow culture score and 'emerging culture' score to cluster together and seeing how many groups 'naturally' emerge'. As far as I understand it, occupational title did not enter into the group of variables used to establish the classes. However, information on occupation is present in the data set so it is possible to see which occupations fall into which classes. But the classificatory criteria really are at the household level. This is different from any other social classification I know of, as all of these are based on characteristics of occupations one way or another. It would have been interesting to see a bit more information about the occupations whose members fell into several different Great British classes. Presumably it would be possible for an office worker, for example, to be in the precariat if they had no house and no savings mixed mostly with people in low status jobs and never went to the opera, and in the established middle class if they had greater income and wealth and different cultural tastes and friends.
Most interesting to me was to see that not all the other variables had a simple graded relationship to income. Table 6 shows an interesting set of relationships. For example, the 'technical middle class' has 3rd highest income and house price, but 2nd highest savings, and lowest frequency of social contacts. Traditional working class households have a higher score on highbrow culture than do 'new affluent workers' or the 'technical middle class'. Presumably this is one of the reasons why people tell me you can change your social class by, for example, saying you like a different kind of music (Morten Wahrendorf offered to lend me some jazz records).
I hope that, for now, this blog has raised some questions in people's minds.
Anyone who read my last blog will also anticipate my reaction to using the term 'social class' to refer to cultural dimensions of social inequality. I just think it is so much less confusing to stick to calling social class what the official UK Statistics Office has defined and validated, i.e. employment relations and conditions.
But OK, let's call the new measure a measure of social position, the general term I like to use to refer to all dimensions of inequality whether they be cultural, economic or occupational.
It is great that the authors start out from a sound knowledge of Goldthorpe and colleagues clear definition and measurement of social class which has provided the present official UK measure (the NS-SEC). Why after all this do they think we need another measure?
Firstly, the NS-SEC does badly in predicting cultural activities and identities. The simple answer to this is that it is not supposed to. Whether or not a class measure based on occupation is related to cultural identity is an empirical matter. Secondly, the NS-SEC does not tell us who is in the 'elite'. I think anyone who uses it would agree with this. Thirdly, economists don't like the NS-SEC. I can surely attest to the truth of this: no record of occupation was recorded in the whole of the English Longitudinal Study of Aging (ELSA) life-grid while the sociologists took their eye off the design. Economists only think of income and wealth, not social class, which is fine, that is their discipline. Another criticism of the Erikson-Goldthorpe-Portocarrero (EGP) class schema (from which the NS-SEC is derived but in a rather distant form) is that it overdoes the manual-non-manual divide. I think John Goldthorpe would agree with this, I certainly do, and it is one reason why the NS-SEC does not have a manual-non-manual divide at all.
Perhaps the ultimate irony of the study is that the authors actually used the NS-SEC to validate their sample distribution! (page 6 Table 1). And this is the correct schema too, not the old EGP, and so it does not have any 'manual workers', or indeed 'skilled or 'unskilled' workers'. Managers are divided into 3 classes and professionals into 2. On the basis that the BBC audience survey (161,000 people) was biased (compared to what I wonder?) the BBC did a face to face survey with a sample of 1026 chosen to be nationally representative. It was on this representative sample that the authors carried out their multiple classification analysis of cultural capital rather than the much bigger BBC audience sample.
Seven classes emerge from the latent class analysis, the result of allowing income, wealth, house price, number and average social prestige of social contacts, highbrow culture score and 'emerging culture' score to cluster together and seeing how many groups 'naturally' emerge'. As far as I understand it, occupational title did not enter into the group of variables used to establish the classes. However, information on occupation is present in the data set so it is possible to see which occupations fall into which classes. But the classificatory criteria really are at the household level. This is different from any other social classification I know of, as all of these are based on characteristics of occupations one way or another. It would have been interesting to see a bit more information about the occupations whose members fell into several different Great British classes. Presumably it would be possible for an office worker, for example, to be in the precariat if they had no house and no savings mixed mostly with people in low status jobs and never went to the opera, and in the established middle class if they had greater income and wealth and different cultural tastes and friends.
Most interesting to me was to see that not all the other variables had a simple graded relationship to income. Table 6 shows an interesting set of relationships. For example, the 'technical middle class' has 3rd highest income and house price, but 2nd highest savings, and lowest frequency of social contacts. Traditional working class households have a higher score on highbrow culture than do 'new affluent workers' or the 'technical middle class'. Presumably this is one of the reasons why people tell me you can change your social class by, for example, saying you like a different kind of music (Morten Wahrendorf offered to lend me some jazz records).
I hope that, for now, this blog has raised some questions in people's minds.
Monday, 11 March 2013
Measuring social inequality
Some time ago I decided I would write something short and sweet about different ways of measuring social inequality. There is a chapter on this in my book on health inequality and although the book is now 10 years old in this respect little has changed. Rather disappointing in fact as we had hoped that work by us and others would soon change the ways in which inequality was measured in health studies. Not to be!
I and my colleagues still get endless papers to review that use the term 'SES' ('socio-economic status') despite this practice having been soundly criticised by Nancy Kreiger in the 1990s, and despite having taken this critique forward in our own empirical research.
What was Nancy's main point? Very simple: it is not helpful in aetiological research to mix up 3 different types of exposure: the social (class) the economic (income) and the cultural (prestige or status). It is a bit like constructing a latent variable made up of smoking, high fat diet and hours of TV watching. These factors will be found to correlate well and to predict health outcomes, but in terms of understanding how the social becomes the biological we will not be any further forward. Rather as if public health had stopped with John Snow and not bothered with microbes, or then had a category 'microbe' covering all different kinds.
What we have tried to do is use the term 'socio-economic position' (SEP) as the most general term to refer to measures of social inequality. And lets be clear here that I mean inequality between individuals not between regions, nations etc. I also get not a negligible number of papers to review that actually think Richard Wilkinson's work is something to do with health inequality (as opposed to health differences between geo-political units with different levels of income inequality). The next level down from SEP is made up of three measures: occupational social class, income, and prestige. The most carefully constructed and well validated measure of social class in use in the UK at present is known as the NS-SEC (National Statistics Socio-Economic Classification). This schema classifies occupations according to clear criteria such as whether a person is self employed or an employee, if self employed how many people they employ, if an employee the presence or absence of career opportunities, discretion over ones own work pattern, responsibility for other people's work content, and job security, summed up as "Employment relations and conditions".
Employment relations and conditions can be classified in the same way in all societies, they do not depend on local cultures. Prestige scales differ between societies and cultures. The Hindu caste system is often taught (or it was when I studied sociology) as the clearest case of a prestige system. Only people of similar caste may worship together, eat together and inter-marry. In the USA occupations have been ranked by prestige by presenting a panel of citizens with a list of occupations and asking them to assign a rank. One obvious difference between caste and the US prestige ratings is that in a caste system your prestige can emanate from your parents, caste is inherited regardless of one's occupation. So someone born into a scheduled caste (the lowest prestige) will have trouble being accepted even if they become a millionaire factory owner. I saw a similar phenomenon here in the UK when it was remarked disparageingly that the mother of Catherine Middleton had been an air stewardess. Of course we do not have caste in the UK! It is just that you can joke about the wife of the future king by calling her mother 'doors to manual', and most people will get it.
One of the sad things about mixing these dimensions on inequlity up (income is obvious so I wont elaborate) is that employment relations and prestige are accompanied by aspects of daily life that have very plausible pathways to health. Castes have different practices in terms of diet for example. Smoking has become a prestige marker in UK but not in Italy or Greece. And so on. I also wonder whether the present debate on 'micro-classes' mixes up these dimensions to some extent and creates problems for research that we don't really need to have if we just keep dimensions separate in the first place.
But this has gone on quite long enough and I will leave it to other people to let me know what they think.
I and my colleagues still get endless papers to review that use the term 'SES' ('socio-economic status') despite this practice having been soundly criticised by Nancy Kreiger in the 1990s, and despite having taken this critique forward in our own empirical research.
What was Nancy's main point? Very simple: it is not helpful in aetiological research to mix up 3 different types of exposure: the social (class) the economic (income) and the cultural (prestige or status). It is a bit like constructing a latent variable made up of smoking, high fat diet and hours of TV watching. These factors will be found to correlate well and to predict health outcomes, but in terms of understanding how the social becomes the biological we will not be any further forward. Rather as if public health had stopped with John Snow and not bothered with microbes, or then had a category 'microbe' covering all different kinds.
What we have tried to do is use the term 'socio-economic position' (SEP) as the most general term to refer to measures of social inequality. And lets be clear here that I mean inequality between individuals not between regions, nations etc. I also get not a negligible number of papers to review that actually think Richard Wilkinson's work is something to do with health inequality (as opposed to health differences between geo-political units with different levels of income inequality). The next level down from SEP is made up of three measures: occupational social class, income, and prestige. The most carefully constructed and well validated measure of social class in use in the UK at present is known as the NS-SEC (National Statistics Socio-Economic Classification). This schema classifies occupations according to clear criteria such as whether a person is self employed or an employee, if self employed how many people they employ, if an employee the presence or absence of career opportunities, discretion over ones own work pattern, responsibility for other people's work content, and job security, summed up as "Employment relations and conditions".
Employment relations and conditions can be classified in the same way in all societies, they do not depend on local cultures. Prestige scales differ between societies and cultures. The Hindu caste system is often taught (or it was when I studied sociology) as the clearest case of a prestige system. Only people of similar caste may worship together, eat together and inter-marry. In the USA occupations have been ranked by prestige by presenting a panel of citizens with a list of occupations and asking them to assign a rank. One obvious difference between caste and the US prestige ratings is that in a caste system your prestige can emanate from your parents, caste is inherited regardless of one's occupation. So someone born into a scheduled caste (the lowest prestige) will have trouble being accepted even if they become a millionaire factory owner. I saw a similar phenomenon here in the UK when it was remarked disparageingly that the mother of Catherine Middleton had been an air stewardess. Of course we do not have caste in the UK! It is just that you can joke about the wife of the future king by calling her mother 'doors to manual', and most people will get it.
One of the sad things about mixing these dimensions on inequlity up (income is obvious so I wont elaborate) is that employment relations and prestige are accompanied by aspects of daily life that have very plausible pathways to health. Castes have different practices in terms of diet for example. Smoking has become a prestige marker in UK but not in Italy or Greece. And so on. I also wonder whether the present debate on 'micro-classes' mixes up these dimensions to some extent and creates problems for research that we don't really need to have if we just keep dimensions separate in the first place.
But this has gone on quite long enough and I will leave it to other people to let me know what they think.
Monday, 18 February 2013
Heritability debates
It is time to try and get to grips with what is happening in genetics/epidemiology. I tend to rely on Meena Kumari, the ICLS expert, but really should get a bit more self-sufficient before I start to write my book. For a long time I was happy that a lot of things which "look genetic" could be, as done elegantly by Debbe Lawlor, alternatively seen as evidence of the huge power of the life course. People are not just born into a home environment, they are to a large extent born into a life course. As Debbie and George Davey Smith showed awhile ago, someone with a high blood level of Vitamin A might not be healthy because of the vitamin but because having a high level of the vitamin was a 'biological embedding' of a whole string of favorable circumstances stretching over years of their lives. Similarly Hilary Graham showed that being a smoker was associated in women with a string of unfavorable experiences in a similar way. I always thought these were profound insights and these papers have greatly influenced my thinking.
So I was not too surprised by the revelations from Genome Wide Association Studies around 2009-10 that only a very small amount of the similarity between people in a lot of characteristics seemed to be explicable in terms of "a gene" (or gene variant, all corrections gladly received). I must admit that only 4% of variation in height attributable to genetics did seem rather small. And if someone had done it for red hair or green eyes, that would have surprised me even more. I don't think anyone did this as hair and eye colour do not carry the same moral/ideological weight as height. But the shock was due to the wide difference between what GWAS studies showed and the estimates of heritability obtained from twin studies, which had been the classic method for assessing the influence of 'genes v. ennvironment' before the genome was sequenced and GWAS became possible.
When people do GWAS (Genome Wide Association Studies), it has 2 phases. First you take a 'training sample'. This has to be very large, and one looks for an association of the characteristic of interest (lets say height) with a Single Nucleotide Polymorphism (SNP) given that it is now possible to include vast numbers of these on a 'SNIP-CHIP'. Because you have no hypothesis about which SNP will be associated with (e.g.) height, you need to get an enormously significant association, as I was once told "p with many many zeroes". The second phase is to take this SNP and see what proportion of the variation in (e.g.) height it explains in a new sample independent of the training sample, the "validation sample". When you do this for height you can explain around 4% of the variation in height. This was quite a shock.
But lets face it what everyone cares about is really how heritable are other personal characteristics with ideological meaning such as intelligence or mental health. As intelligence and mental health are generally acknowledged to be complex characteristics, much more than height, what hope was there?
But wait! Along came Professor Visscher and colleagues. Their paper in Twin Research and Human Genetics (vol 13(6) pp. 517-24) aims to set out their method in a way that is more comprehensible than their original paper in Nature Genetics that shows after all about 40% of height is explained by genetics. Never mind, they said, if there is no individual SNP associated significantly with height that explained more than 4% of it. What you need to do is to add up all the SNPs that are associated with height, but so weakly that they do not meet the very stringent requirements for significance used in GWAS studies. This can then be extended to intelligence with results that are summarised as follows:
So I was not too surprised by the revelations from Genome Wide Association Studies around 2009-10 that only a very small amount of the similarity between people in a lot of characteristics seemed to be explicable in terms of "a gene" (or gene variant, all corrections gladly received). I must admit that only 4% of variation in height attributable to genetics did seem rather small. And if someone had done it for red hair or green eyes, that would have surprised me even more. I don't think anyone did this as hair and eye colour do not carry the same moral/ideological weight as height. But the shock was due to the wide difference between what GWAS studies showed and the estimates of heritability obtained from twin studies, which had been the classic method for assessing the influence of 'genes v. ennvironment' before the genome was sequenced and GWAS became possible.
When people do GWAS (Genome Wide Association Studies), it has 2 phases. First you take a 'training sample'. This has to be very large, and one looks for an association of the characteristic of interest (lets say height) with a Single Nucleotide Polymorphism (SNP) given that it is now possible to include vast numbers of these on a 'SNIP-CHIP'. Because you have no hypothesis about which SNP will be associated with (e.g.) height, you need to get an enormously significant association, as I was once told "p with many many zeroes". The second phase is to take this SNP and see what proportion of the variation in (e.g.) height it explains in a new sample independent of the training sample, the "validation sample". When you do this for height you can explain around 4% of the variation in height. This was quite a shock.
But lets face it what everyone cares about is really how heritable are other personal characteristics with ideological meaning such as intelligence or mental health. As intelligence and mental health are generally acknowledged to be complex characteristics, much more than height, what hope was there?
But wait! Along came Professor Visscher and colleagues. Their paper in Twin Research and Human Genetics (vol 13(6) pp. 517-24) aims to set out their method in a way that is more comprehensible than their original paper in Nature Genetics that shows after all about 40% of height is explained by genetics. Never mind, they said, if there is no individual SNP associated significantly with height that explained more than 4% of it. What you need to do is to add up all the SNPs that are associated with height, but so weakly that they do not meet the very stringent requirements for significance used in GWAS studies. This can then be extended to intelligence with results that are summarised as follows:
Data from twin and family studies are consistent with a high heritability of intelligence, but this inference has been controversial. We conducted a genome-wide analysis of 3511 unrelated adults with data on 549 692 single nucleotide polymorphisms (SNPs) and detailed phenotypes on cognitive traits. We estimate that 40% of the variation in crystallized-type intelligence and 51% of the variation in fluid-type intelligence between individuals is accounted for by linkage disequilibrium between genotyped common SNP markers and unknown causal variants. These estimates provide lower bounds for the narrow-sense heritability of the traits. We partitioned genetic variation on individual chromosomes and found that, on average, longer chromosomes explain more variation. Finally, using just SNP data we predicted ~1% of the variance of crystallized and fluid cognitive phenotypes in an independent sample (P=0.009 and 0.028, respectively). Our results unequivocally confirm that a substantial proportion of individual differences in human intelligence is due to genetic variation, and are consistent with many genes of small effects underlying the additive genetic influences on intelligence. [my emphasis]G Davies, A Tenesa, A Payton, J Yang, S E Harris, D Liewald, X Ke, S Le Hellard, A Christoforou, M Luciano, K McGhee, L Lopez, A J Gow, J Corley, P Redmond, H C Fox, P Haggarty, L J Whalley, G McNeill, M E Goddard, T Espeseth, A J Lundervold, I Reinvang, A Pickles, V M Steen, W Ollier, D J Porteous, M Horan, J M Starr, N Pendleton, P M Visscher and I J Deary "Genome-wide association studies establish that human intelligence is highly heritable and polygenic" Molecular Psychiatry (2011) 16, 996–1005;
However, Makowsky et al. for example threw some doubt on the enterprise in PLOS Genetics thus:
Recently, a large proportion of the “missing heritability” for human height was statistically explained by modeling thousands of single nucleotide polymorphisms concurrently. However, it is currently unclear how gains in explained genetic variance will translate to the prediction of yet-to-be observed phenotypes. Using data from the Framingham Heart Study, we explore the genomic prediction of human height in training and validation samples while varying the statistical approach used, the number of SNPs included in the model, the validation scheme, and the number of subjects used to train the model. In our training datasets, we are able to explain a large proportion of the variation in height (h2 up to 0.83, R2 up to 0.96). However, the proportion of variance accounted for in validation samples is much smaller (ranging from 0.15 to 0.36 depending on the degree of familial information used in the training dataset). While such R2 values vastly exceed what has been previously reported using a reduced number of pre-selected markers (<0.10), given the heritability of the trait (~0.80), substantial room for improvement remains. Ana I. Vazquez,Christine W. Duarte,David B. Allison, Gustavo de los Campos. Beyond Missing Heritability: Prediction of Complex Traits. PLOS Genetics 2011; 7(4); e1002051The furthest I can get today is to comment on this phrase "substantial room for improvement". It sounds as if people have already decided what % of the trait is "heritable" and are now trying to find a method that gives the answer they already expect and want to find.
Sunday, 20 January 2013
God need not bring data
The International J of Epidemiology for December has a fascinating paper by Lynch and Stuckler on the use of data for quality control by Deming and its implications for public health. I thought people might like to see some of what the authors say about data.
So I warmly welcome the initiative of the IJE to bring together health data sets as a public, openly shared resource. I would hope they will add the UK Data Archive to their list as it now contains some biomedical and genetic data and will soon have more.
I hope that Deming would approve of this method for data curation and exploitation. It means that work done using taxpayers' money (ESRC funded project are obliged to archive data) become a common good for the whole academic and policy community. Also, in the words of an eminent colleague: "if you are not allowed to see the data behind a paper how do you know it is not all made up?". So open data is a vital safeguard against the kind of scientific misconduct that is increasingly being noted.
A number of ruses get used to try and avoid 'sharing' data (another eminent colleague tells me not to use this term as the data is not the property of the research team in the first place). I have heard it said that 'biomedical data (like blood pressure and cholesterol) carry a bigger risk of disclosure' and some social scineitists are fooled by this. Maybe they are thinking about "CSI" on the TV. In fact data on occupation and education are a lot more potentially disclosive that biomedical markers. I also hear it said that biomedical samples are a 'depletable resource'. But what gets archived is just a bunch of codes, not the samples themselves! I once heard it said that the low response rate of studies like the BioBank (10% response rate, i.e. 90% non-response) "does not matter as it is only going to be used for case-control studies". I will leave other people who know more epidemiology than me to react to this statement but the ones I speak to just roll their eyes.
"Available data often go unused because they are not well enough documented, lack accessible how-to guides for their use, or knowledge about the resource is passed on informally within research groups or collaborations. Some data may also require analytical skills that are in short supply; or people may simply be unaware of their existence or unable to access them."This may sound familiar to some readers. The research group around ICLS uses only openly available data, not the kind that is not well enough documented, lack accessible how-to guides for their use, or knowledge about the resource is passed on informally within research groups or collaborations. We do this to protect ourselves from delays in data acquisition, people changing their minds about whether or not we are 'allowed' to use data for a specific purpose, and from having to re-code things that have not been through the quality control of the UK Data Archive and a variety of users. When we find mistakes or derive new variables we give the code to the Archive (access to the data is free to anyone funded by UK Research Councils).
So I warmly welcome the initiative of the IJE to bring together health data sets as a public, openly shared resource. I would hope they will add the UK Data Archive to their list as it now contains some biomedical and genetic data and will soon have more.
I hope that Deming would approve of this method for data curation and exploitation. It means that work done using taxpayers' money (ESRC funded project are obliged to archive data) become a common good for the whole academic and policy community. Also, in the words of an eminent colleague: "if you are not allowed to see the data behind a paper how do you know it is not all made up?". So open data is a vital safeguard against the kind of scientific misconduct that is increasingly being noted.
A number of ruses get used to try and avoid 'sharing' data (another eminent colleague tells me not to use this term as the data is not the property of the research team in the first place). I have heard it said that 'biomedical data (like blood pressure and cholesterol) carry a bigger risk of disclosure' and some social scineitists are fooled by this. Maybe they are thinking about "CSI" on the TV. In fact data on occupation and education are a lot more potentially disclosive that biomedical markers. I also hear it said that biomedical samples are a 'depletable resource'. But what gets archived is just a bunch of codes, not the samples themselves! I once heard it said that the low response rate of studies like the BioBank (10% response rate, i.e. 90% non-response) "does not matter as it is only going to be used for case-control studies". I will leave other people who know more epidemiology than me to react to this statement but the ones I speak to just roll their eyes.
Subscribe to:
Posts (Atom)