Skip to main content
. 2007 Nov 30;13(4):193–202. doi: 10.2188/jea.13.193

Figure 1. Boxplots exhibiting 100 simulated estimates of the multiplicative interaction parameter γM (equation 2) and 95% likelihood-based confidence bounds for the three sampling methods: matched (M), counter-matched (C), and unmatched (U).
Estimates of the parameter and its confidence bounds were obtained by exponentiation of those for γM. Results are for case-control ratios of 1:1, 1:2, 1:3, 1:5, and 1:9. Results for counter matching with two controls are from the design with the extra control unexposed. Narrower boxplots reflect smaller sampling variation in the interaction estimates or confidence bounds. More precise estimation of interaction is evidenced by median bounds closer to the interaction estimate. For comparison, the cohort interaction parameter (1.14) and confidence limits (0.74, 1.73) are shown as dotted lines.

Figure 1.