Testing Causal Explanations in Organismal Biology: Causation, Correlation and Structural Equation Modelling

This paper describes some recent statistical developments that specify under what conditions multivariate causal explanations can be tested with observational data. The specific questions that motivate the paper involve testing hypotheses concerning the causal relationships between different traits of an organism. I argue that in such cases neither randomized experiments nor controlled experimental manipulations can test such hypotheses but that structural equation modelling (path analysis) can be used for such purposes. I conclude by describing some of the limitations of path analysis.