Figure 4.

Bias in combined cases. Simpson's paradox may occur when the fraction of cases that is drug resistant is only counted among combined cases and not separately among new and retreatment cases. By examining public sector combined cases only we overestimate the total fraction of resistant cases in the population, while among both subcategories (new and retreatment cases) we underestimate the fraction that are resistant. Results present values at equilibrium with aN = 0.1; aP = 0.1; fSN = 0.1; fSP = 0.1; fRN = 0.25; fRP = 0.25; l = 0.2; q = 0.6; rN = 0.85; rP = 0.5; xR = 0.16; xS = 0.2

Cohen et al. BMC Public Health 2010 10:355   doi:10.1186/1471-2458-10-355
Download authors' original image