Thanks for visiting Scientific Computing World.

You're trying to access an editorial feature that is only available to logged in, registered users of Scientific Computing World. Registering is completely free, so why not sign up with us?

By registering, as well as being able to browse all content on the site without further interruption, you'll also have the option to receive our magazine (multiple times a year) and our email newsletters.

What are your thoughts?

Share this on social media:

Topic tags: 

Felix Grant explores the ways in which statistical data analysis is vital to psychology

‘I used to think I was indecisive, but now I’m not so sure.’ So runs one of the oldest entries in the bumper fun book of psychology jokes. It parallels one in the statistician’s equivalent volume: ‘Statistics means never having to say you’re certain’ which, if you are too young to remember Ali McGraw and Ryan O’Neal[1] in Love Story, plays off the strapline ‘Love means never having to say you’re sorry.’ And both have serious echoes in professional practice because psychology is an area in which data trees are always compromised by a forest of confounding factors and samples can often be small. In most pure, classical psychology, to an even greater extent than in the physical sciences, only statistical analysis can, to mix metaphors, tease out the pure signal from the white noise with confidence.

Even in its pure and classic form, the concerns of even the most abstract psychological research are rarely far from pragmatic utility. Roland Bremond, a mathematical morphologist currently concerned with the very down-to-earth business of transport planning, neatly encapsulated the data analytic nature of the beast[2] in Saussurian terms: ‘La “réalité”, du monde des objets, en psychologie expérimentale, est définie statistiquement comme l’ensemble des propriétés perçues partagées par tous les sujets “normaux”... Autrement dit, la réalité est une propriété statistique, sur une population...’ (In experimental psychology ‘reality’, the world of objects, is defined statistically as the set of properties shared by all perceived ‘normal’ subjects... or, to put it another way, reality is a statistical property of a population...)

Not that pure classical psychological research is the only strand or even the most evident. Without straying too far into monist and dualist controversies, it’s fair to say that complete separation of mind from body was abandoned a long time ago and psychology interpenetrates deeply with neuroscience and biochemistry amongst many other physical areas.

In its classic guise, especially in combination with modern computerised statistics, psychology is one of the areas of science where real, useful, original research can still be done by the solitary researcher without institutional funding or resources. In my freelance consultancy life I regularly work with research psychologists and frequently encounter such individuals, most of them right at home in one data analysis software package or another. In the run up to this article, several lines of thought came from lone workers including the occasional pre-university student.

Taking Bremond’s transport interest as a departure point, one motivator for turning to psychology is the very high potential cost (economic and human) of error in areas where people interact with kinetic technologies. Psychological assessment of pilots and other crew is a well-established contributor to the enviable safety record which airlines enjoy in relation to other modes of travel. Drivers of road vehicles are not, for obvious reasons to do with volume feasibility, monitored in the same way and investigation of how their psychological states affect behaviour can only, realistically, ever be statistical.

A look at the literature shows that this is reflected in the different types of research which predominate. Airline and air traffic control studies have an emphasis on understanding conceptual issues such as, to take an example[3] which happens to be on my desk, the differing cognitive states which underlie modes of professional attention. Research into traffic psychology tends, by contrast, to seek predictors for particular behaviours associated with increased risk taking by individual drivers.  So, for example, a quick opportunity sample of papers on my desk show intensely data analytic investigations into factors which predict aggressive road behaviour[4], associations between of attention deficit symptoms to driver decision making[5], psychological linkages between degree of multitasking and concentration on priority foci[6], factors affecting perception of motorcyclists by other road users[7], accuracy or otherwise in assessment of own visual acuity in adverse conditions[8], risk perception differences between adult and adolescent drivers, and so on. I asked transport psychology researchers from each school what statistical software they used and both said Systat, but while the one studying air traffic controllers said that he rarely used it for anything more than basic description, his opposite number concerned with motorway emergency management brought out analyses across most of the range of which the package is capable.

As with most knowledge, there are few if any impermeable membranes separating results of psychological research. What inclines someone to be an aggressive driver, for instance, may well influence their likelihood of aggression in other areas too. Several researchers I spoke to were working within, or in association with, police forces and however diverse their areas of study, their measures and methods were very similar.

Sonja, funded by a north European law enforcement agency to study the psychological indicators of adolescents who do and do not join nascent gang cultures, finds many areas of overlap with the aggressive driving researchers. So does Anton who, further east, tries through analysis of recorded custody interviews, to make statistical sense of psychological patterns in racial and religious violence. Sonja and Anton exchange data and findings with one another on an unofficial ad hoc level; one runs intensive batch processing in R, the other in Statistica, and there is a language barrier but they use SigmaPlot as a lingua franca and enthuse about the synergistic value of the collaboration.

Moving back a little from such concrete concerns to more fundamental quests for understanding, and also a little way into the borderlands of neurobiology, a study of emotional response by Martin Klasen (University of Aachen) and others[9] also illustrates multiple scientific computing roles in experimental psychology.

Studying the supramodal representation of emotion, Klasen used virtual reality technology (thus avoiding the difficulty for a human actor of trying to present incongruent signal combinations) to generate avatars displaying a range of facial expressions representing different emotional states. Spoken disyllabic pseudo words with the patterns of stress and intonation corresponding to similar states were lip synchronised to the avatars, sometimes congruently and sometimes not (48 combinations in each case) in a block experimental design mixing audio, visual and audio-visual versions. Respondents were asked to quickly classify the results across nearly 400 individual trials.

From analysed functional magnetic resonance imaging (fMRI) scans, a multiple predictor random-effects general linear model was then used to generate statistical parametric maps according to stimulus type. Mean testing and repeated-measures ANOVA testing were applied to the response data alongside conjunctional analysis of the fMRI results. The outcome, cutting to the chase, was a strong indication that visual and audible signals are emotionally integrated late in the limbic system’s software processing rather than at the perceptual level.

One of my pre-university correspondents, prompted by experience of juggling exams and the 24-hours-a-day demands of a young baby, is conducting an impressively rigorous piece of primary research on sleep deprivation in adolescent parents, running data analyses in Mystat, the free student version of Systat.

Sleep deprivation seems to be associated in most people’s minds with deliberate use as a dislocator in interrogation, but is in reality widespread as a result and cause of medical problems across the general population. The mechanisms are sometimes purely physical but far more often psychological or psychosomatic; they are highlighted for instance in a report[10] on holistic approaches to critical care.

A paper[11] published in the British Medical Journal earlier this year illustrates the vicious circle. Patients recovering from surgical intervention in a life-threatening condition suffer depression and anxiety; this disrupts sleep; loss of sleep leads to pain, fatigue and reduced ability to concentrate or deal with routine tasks, which in turn damage recuperation. This study is concerned with ameliorative melatonin treatment, but Asma, a GP with a large number of patients in a socially deprived area of a European conurbation for whom chronic sleep disruption is a common factor in such recuperation problems, uses free online statistical analysis tools, such as those for nonparametric testing at Vassar College’s website, to refine non-pharmaceutical approaches to breaking similar cycles.

Though it is a natural extension of the same thinking as the sleep deprivation issue, I was separately set off along the psychosomatic track by an email enquiry about statistical methods. Twenty-year-old student Aqib Ashraf is interested in the association between psychological state and resistance to viral infection in his peers, and wants to design his research with sound data analysis in mind.

There is a lot of assumption and anecdotal report in this area, but some quantitative work has also been done. Sheldon Cohen of Carnegie Mellon University’s Department of Psychology, who has, with various collaborators, pursued this line over a period of 20 years or so, shows that subjects show a statistically significant relation between positive emotional style (PES) and not only reduced reporting of symptoms, but objectively verifiable rates of rhinovirus and influenza infection. Cohen concludes[12] on the basis of a trial with almost 200 subjects that ‘positive emotions play a larger and more important role in disease risk and health complaints than previously believed.’

Taking a longer and wider view of linkage between psychological states and bodily wellbeing, Schafer and Ferraro (University of Toronto and Purdue University, respectively) conducted a multivariate statistical analysis of longitudinal data from just under 3,000 adults. Their results show[13] a high degree of correspondence between incidence of declared childhood stressors, such as abuse or financial strain, and probability of a range of physical illnesses later in life. After allowing for confounding factors, there is no statistically sustainable evidence of influence on levels of cancers, strokes or heart problems, but 15 identified linkages with probabilities below five per cent include conditions as diverse as haemorrhoids, sciatica and emphysema. The authors draw an approximate equivalence between the aggregate psychosomatic effects of childhood misfortune on adult disease avoidance and ‘the combined effect of moderate lifetime smoking and obesity.’

I opened by saying that statistical data analysis is vital to psychology, but the traffic is not one way. The psychology of data producers, users and consumers is an inescapable factor in the manner and quality of analytic process, which in turn affects the scientific validity and value. This realisation has (see box: Psyching out the data) strongly shaped the presentation of data and lies behind the evolution of exploratory data analysis which so often opens up avenues of research for more formal and traditional methods.

References and Sources

For a full list of references and sources cited in this article, visit www.scientific-computing.com/features/referencesaug12.php

Psyching out the data

Leland Wilkinson, originator of the Systat data analysis and visualisation software, and of Java analytical graphics foundation nVIZn, has long and variously argued that graphical data visualisations need to be conceived and designed within a transactional grammar shaped by the psychology of human perception.

‘We map in order to organise our world in our mind,’ says Wilkinson[14], ‘...abstract reasoning is built on metaphors for reality.’ He has devoted a significant proportion of his career to developing visualisation techniques founded on this approach, strongly influencing statistical graphics thinking in general. Compared to its market sector as a whole, Systat contains a very high proportion of plotting methods based on psychoperceptual principles, pulling in those developed by others as well as by Wilkinson himself. He has recently launched a new exploratory system built on his precepts.

The use of psychological thinking, whether explicit or implicit, in designing data communication is of course older than scientific computing. Wilkinson mentions[14] Charles Joseph Minard’s compound graphic illustrating French losses in the 1812-1813 Russian campaign, and Florence Nightingale’s use of polar area ‘rose diagrams’ in the 1850s is another well-known example. Nevertheless, it is the use of scientific computing methods that has allowed this approach to flower as a widespread means of data analytic communication.

The most visible current expression of this is in journalistic encapsulation of complex situations by broadsheets, television news and websites – Guardian Data being a well-known user. Tableau Software, whose frequent recruitment advertisements for psychology graduates are a clear indicator of their approach, are behind many well-known examples of this trend, providing both commercial and no cost software with web platforms upon which the results can be placed. Drag and drop variables entry, multiple views onto the same data, sophisticated data brushing to provide simultaneous subsetting in all of those views and use of multiple modalities all play to the inherent mapping faculties of human psychology.

‘Colour is a great way to encode categorical data for visual analysis because the human visual system can isolate marks for visual comparison,’ comments Jock McKinley, director of Visual Analysis at Tableau. ‘However, as number of categories grows, the colours can become similar, which makes it hard for the visual system to isolate a specific colour... [and]... we recommend that you encode the data with shape... perceptual best practices like these help people to achieve speed-of-thought analysis.’