The distribution of the cross product ratio in the bivariate log normal distribution, with an application to non-randomised trials

Marchant, P. R., & Baxter, P. D.

Non-randomised trials can give a false impression of the effectiveness of any intervention. We outline and illustrate a method to estimate the size of the bias under the assumption of bivariate normality of log incidence rates.

Keywords: bivariate log normal distribution, crime reduction interventions, non-randomised trials, regression towards the mean