Skip to main content
. Author manuscript; available in PMC: 2019 Jan 1.
Published in final edited form as: Epidemiology. 2018 Jan;29(1):58–66. doi: 10.1097/EDE.0000000000000765

Figure 2.

Figure 2

Relative Variance across 250 replicates as a function of data features for the severe chronic obstructive pulmonary disease outcome: We show RV=Var^¯RS(β^)/Var^¯AVS:SOR(β^) for a number of data features. Black diamonds denote the analyses presented in Table 3 and grey diamonds show the result of perturbing one data, analysis or design feature. Panels show the impact on relative variance of response prevalence (A), strength of the Z ~ Y relationship (B), amount of response dependence (C), richness of the auxiliary variable model for Z (D), and asthma exposure and auxiliary variable sampling (E).