FIGURE 1.

An example of individual‐level Monte Carlo covariate simulation where the joint distribution of three baseline characteristics, age, comorbidity c and comorbidity d, is factorized into the product of marginal and conditional distributions, such that p(age, c, d) = p(d∣c, age)p(c∣age)p(age). The joint distribution is valid because the conditional distributions defining the covariates are compatible: we start with a marginal distribution for age and construct the joint distribution by modeling each additional covariate, one‐by‐one, conditionally on the covariates that have already been simulated. This diagram adopts the convention of Kruschke. ⁴⁷ [Colour figure can be viewed at wileyonlinelibrary.com]