From conventional hypothesis testing in science to the evolutionary hypothesis
Author: Miguel Ángel Martínez, Full Professor of Preventive Medicine and expert in Biostatistics at the University of Navarra.
Date of publication: Pamplona, December 21, 2011
The concept of causality is central to public health and preventive medicine. Without addressing the causes of disease, effective prevention is not possible. Modern epidemiology has developed mainly by dealing with the search for the causes of health and disease. An indispensable tool in this endeavour is biostatistics.
Biostatistics makes extensive use of hypothesis testing. In theory, chance could explain any association or phenomenon in nature, no matter how strange or improbable it might seem. But biostatistics uses hypothesis testing to estimate the probability of finding a result at least as strange as the observed one, if everything were due merely to chance. Conventionally, when such a probability is leave, less than 0.05, the hypothesis that chance could explain everything is rejected and a statistically significant association is claimed instead. These last two words have plagued the current scientific literature, perhaps with the exception of evolutionary biology, where they do not seem to be so abundant. But a mere association, however significant, does not prove causality.
Epidemiology goes a step further by using a series of criteria and models to move from the "statistically significantassociation " to a true cause-effect relationship. In these models, the fraction attributable to causation is opposed to the fraction attributable to chance. Nothing could be further from my pretensions than to deny that there is any subject of evolution, but it is at least intriguing to an epidemiologist that radical evolutionism seems to equate chance with causation.