Table 2.

Description of the performance‐settings and corresponding test datasets

Performance‐setting	Description	Test set data‐generating mechanism
Mix of Treatment	Model validation on samples	Test set 1 (N = 100 000):
(MT)	drawn from a similar	Generated under exactly the same
	population to the	process as the development cohort.
	development set.
	Corresponds to estimating E1.
No Baseline Treatment	Model validation on samples	Test set 1 (N = P(A0 = 0)×
(NBT)	drawn from a similar	100 000):
	population to the	Generated under exactly the same
	development set, but	process as the development cohort,
	restricted to those without	but restricted to examining
	treatment at baseline.	{i ∈ [1, N] : a0i = 0}.
	Corresponds to estimating E2.
No Treatment	Model validation in a	Test set 2 (N = 100 000):
Throughout (NTT)	population where treatment is	generated as
	withheld from all patients,	X0 ∼ N(0, 1)
	but where the distribution of	X1 ∼ N(X0, 1)
	covariates is similar to the	A0 = A1 = 0
	development cohort.
	Corresponds to estimating E3.