Table 1.

Estimated Log Relative Rate per 100 WLMs Under Various Approaches to Selecting the Exposure Lag Assumption^a

Three Scenarios Regarding σw	Estimated β	Empirical SE	Estimated SE	CI Coverage
σw = 0.1
Select “guessed” lag that yields largest estimate of association	1.33	0.126	0.120	0.218
Select “guessed” lag that yields best model fit	1.04	0.195	0.092	0.623
Simultaneously estimate lag and estimate of association	1.01	0.190	0.134	0.878
σw = 0.5
Select “guessed” lag that yields largest estimate of association	1.31	0.095	0.093	0.083
Select “guessed” lag that yields best model fit	1.03	0.147	0.072	0.684
Simultaneously estimate lag and estimate of association	0.99	0.114	0.077	0.880
σw = 1.0
Select “guessed” lag that yields largest estimate of association	1.24	0.076	0.070	0.050
Select “guessed” lag that yields best model fit	1.02	0.078	0.056	0.851
Simultaneously estimate lag and estimate of association	1.01	0.071	0.055	0.915

Abbreviations: CI, confidence interval; SE, standard error; WLM, working level month.

^aIn all simulations, the specified true association is β = 1. The natural log of annual exposure is distributed with μ = 0.1 and σ_B = 1.