Content area
Full Text
In a recent paper, Messerli 1 reported a strong positive association between national chocolate consumption and the number of Nobel prizes a country has won. If this association had occurred at the level of individual people, then the conclusions would have been remarkable, and academics around the world would doubtless be ordering chocolate by the tonne in an attempt to realise their lifelong dreams. His analysis, however, was what epidemiologists call an ecological analysis. Data were aggregated by country, and so the results suggested that countries with high chocolate consumption have tended to produce many Nobel prize winners; crucially, however, we do not know what were the eating habits of the prize winners themselves. Ecological studies are generally regarded as hypothesis-generating; those of us who try to teach the basics of epidemiology to medical students warn them of possible dangers by urging them to 'Beware the ecological fallacy'. Put simply, this says that associations found at aggregate level, whether the aggregation is by country, city or even smaller units, do not necessarily apply to individuals living there. There are some famous examples of this fallacy; in 1930, Robinson 2 studied illiteracy in US states and showed a substantial negative correlation of -0.53 between the percentages of foreign-born residents and illiterate adults when the data were aggregated by state, but an individual-level analysis showed a positive association between being foreign born and being illiterate.