Abstract
I write out and discuss how one might try to prove the continuous
time g-computation formula, in the simplest possible case: treatments
(labelled a, for actions) and covariates (l: longitudinal data) form together
a bivariate counting process.
