Table 2.

Comparison of the log-rank type test and the additive model: power

	62% events due to ex. haz.				93% events due to ex. haz.
	LRt	sAM	fAM	LRh	LRt	sAM	fAM	LRh
X∈D, bal, bin (sex)	0.499	0.561	0.552	0.786	0.878	0.874	0.877	0.903
X∉D, bal, bin	0.487	0.578	0.567	0.792	0.877	0.877	0.876	0.903
X∉D, imbal, bin, ef >0	0.467	0.391	0.359	0.700	0.857	0.793	0.792	0.847
X∉D, imbal, bin, ef <0	0.537	0.720	0.711	0.864	0.888	0.931	0.933	0.943
X∉D, bal, 4 grps	0.339	0.408	0.385	0.629	0.736	0.743	0.745	0.787
X∉D, bal, bin, NPH, ef ≈0	0.052	0.074	0.062	0.047	0.048	0.053	0.051	0.049
X∉D, bal, bin, NPH	0.524	0.510	0.476	0.819	0.880	0.865	0.861	0.912

Methods included: log-rank type (LRt), semi-parametric additive model (sAM), fully parametric additive model (fAM), log-rank test in hypothetical world (LRh). ((im)bal = (im)balanced variable, i.e., the groups occur with (un)equal probabilities; bin= binary variable; 4 grps = a variable with four groups; ef = variable’s effect; NPH = nonproportional effect). X is the categorical covariate of interest, D denotes the demographic variables