Screw you, p-values. More fundamentally, when studying culture and its principal alternatives, ecological variation and genetic causation, there is no consensus as to which direction we should lean, or, more formally, what is the null hypothesis (or ‘simple’ explanation) and what is the alternative (or ‘sophisticated’ explanation). In studies of humans, culture is usually taken as the null, and so the burden of proof lies with those testing hypotheses about genetic causation, while in non-human studies the direction of proof is always reversed, and so ascribing culture without reason would be the false positive. This is highly unsatisfactory. The approach is set up along the lines of addressing ‘the question of culture’7. This implies there must be a yes or no answer. Set up as a dichotomous hypothesis-testing approach, where everything hinges upon getting across a completely arbitrary ‘significance level’ in statistical tests, the approach is almost guaranteed to give incomplete answers. This approach has been much discredited by statisticians themselves, and by scientists working in applied areas who have to make real decisions about, for example, wildlife management8. We therefore favor conceptual frameworks which treat the potential causes of behavioral variation on the same basis, and take a model comparison rather than binary hypothesis testing approach. In other words, the starting point should be that genetic, ecological and social influences are all likely to play some sort of role, and the research effort should be expended in understanding the relative contributions of each, by comparing how well statistical models in which the relative contributions of each are allowed to vary are able to explain what we see in nature9. This seems to us much more likely to bring useful insights
Posted on: Wed, 14 May 2014 01:32:27 +0000
Trending Topics
Recently Viewed Topics
© 2015