Chapter 9 Causality and Mediation

Causality is a concept that statisticians and statistical science traditionally shy away from. Recently, however, many successful attempts have been made to include the concept of causality in the statistical theory and vocabulary. A good review of the topic, from a modern, event history analysis point of view, is given in Aalen, Borgan, and Gjessing (2008), Chapter 9. Some of their interesting ideas are presented here.

The traditional standpoint among statisticians was that “we deal with association and correlation, not causality”, see Pearl (2000) for a discussion. An exception was the clinical trial, and other situations, where randomization could be used as a tool. However, during the last couple of decades, there has been an increasing interest in the possibility to make causal statements even without randomization, that is, in observational studies (Rubin 1974; Robins 1986).

Matching is statistical technique, which has an old history without an explicit connection to causality. However, as we will see, matching is a very important tool in the modern treatment of causality.

Unfortunately, the models for event history analysis presented here are not implemented in R or, to my knowledge, in any other publicly available software. One exception is matched data analysis, which, except the matching itself, can be performed with ordinary stratified Cox regression.

References

Aalen, O. O., Ø. Borgan, and H. K. Gjessing. 2008. Survival and Event History Analysis: A Process Point of View. New York: Springer.

Pearl, J. 2000. Causality: Models, Reasoning and Inference. New York: Cambridge University Press.

Robins, J. M. 1986. “A New Approach to Causal Inference in Mortality Studies with a Sustained Exposure Period—Application to Control of the Healthy Worker Survivor Effect.” Mathematical Modeling 7: 1393–1512.

Rubin, D. B. 1974. “Estimating Causal Effects of Treatments in Randomized and Non-Randomized Studies.” Journal of Educational Psycology 66: 688–701.