Table 2.

Results from estimating a joint model on simulated data based on 200 replications per scenario

			Size Subcohort: 1/3				Size Subcohort: 1/6
% Events		Scenario				Scenario
			FC	CCI	CCII		FC	CCI	CCII
20%	α	1	0.975	0.971	0.849	2	0.976	0.966	0.799
	bias		−0.025	−0.029	−0.151		−0.024	−0.034	−0.201
	(2.5%‐97.5%)		(0.89‐1.07)	(0.88‐1.07)	(0.76‐0.94)		(0.89‐1.07)	(0.88‐1.06)	(0.71‐0.89)
	coverage		92%	91%	13%		92%	88%	4%
	β 1		1.003	0.996	1.087		1.004	0.986	1.139
	β 2		0.319	0.331	0.558		0.324	0.357	0.713
	β 3		0.110	0.104	0.142		0.109	0.097	0.154
	β 4		0.104	0.105	0.092		0.102	0.099	0.092
	γ		−1.979	−1.987	−1.774		−1.979	−1.978	−1.676
5%	α	3	0.856	0.845	0.727	4	0.858	0.835	0.649
	bias		−0.144	−0.155	−0.273		−0.142	−0.165	−0.351
	(2.5%‐97.5%)		(0.74‐0.99)	(0.72‐0.98)	(0.61‐0.86)		(0.74‐0.99)	(0.71‐0.97)	(0.53‐0.78)
	coverage		38%	33%	1%		39%	32%	0%
	β 1		1.003	0.993	1.062		1.005	0.990	1.127
	β 2		0.331	0.343	0.474		0.334	0.371	0.638
	β 3		0.108	0.099	0.127		0.106	0.087	0.146
	β 4		0.101	0.103	0.055		0.100	0.107	0.023
	γ		−2.730	−2.760	−2.421		−2.771	−2.806	−2.238

The bias indicates the difference between the simulated parameter value and the estimated value by each of the models. The coverage is calculated by the percentage of times the true simulated values falls in the credible interval of each simulation.

Simulated values of the parameters: α = 1, β ₁ = 1, β ₂ = 0.3, β ₃ = 0.1, β ₄ = 0.1, γ = −2.

Abbreviations: CCI, case‐cohort design, retain all survival information; CCII, case‐cohort design, classical version; FC, full cohort.