Table 1. Estimates of the parameters used in our dynamic HIV-FGS model (Figure 1).

Variable	Meaning (units)	Prior distribution [ref]	Posterior distribution: median (95% CI)
	Annual growth rate rural population	0.034 [29]	N/A
	Duration of sexual activity (year-1)	1/35	N/A
	Enhanced HIV transmission due to FGS	Uniform(0,20)	5.9 (3.8–9.1)
	Probability of FGS given childhood infection	Uniform(0.33,0.75) [8]	0.47 (0.38–0.56)
	Probability of acquiring FGS during adulthood	Uniform(0.005,0.05)	0.008 (0.006–0.009)
	Duration of HIV infection (years)	Uniform(7.5,12.5) [33, 48]	10.7 (8.1–12.1)
	Number of sex acts in partnerships per year for high-risk group	Uniform(15,150) [48]	128 (95–148)
	Number of sex acts in partnerships per year for low-risk group	Uniform(50,248) [48]	69 (28–137)
	HIV transmission rate per sex act	Uniform(0.0006,0.004) [23], [39]	0.0022 (0.0009–0.003)
	Mixing between sexual risk groups	Uniform(0.2,0.9) [49]	0.44 (0.22–0.62)
	Extent to which males determine the pattern of sexual partnerships formation	Uniform(0.2,0.8) [48]	0.67 (0.50–0.78)
	Initial partner change rate: women (year-1)	Triangular(0.66,2.4,0.9) [48]	1.27 (0.7–2.2)
	Initial partner change rate: men (year-1)	Triangular(1.1,3,1.2) [48]	1.9 (1.2–2.8)
	Fraction of women in high-risk group	Uniform(0.05,0.6) [48]	0.20 (0.15–0.24)
	Fraction of men in high-risk group	Uniform(0.10,0.75) [48]	0.34 (0.22–0.49)
	Relative rate of partner change: high-risk versus low-risk group	Uniform(1,100) [48]	10.7 (2.5–21.8)
	Reduction rate of partner change	Uniform(1,50)	6.5 (3.5–9.0)
	Year HIV epidemic starts	Uniform(1978,1985) [48,50]	1981 (1980–1982)

These parameter estimates produced the best fit of our dynamic model to epidemiological data for HIV and FGS prevalence and co-infection among rural Zimbabwean women [3], [7]. The dynamic model was fit to these data using a Markov Chain Monte Carlo method, which allowed us to calculate distributions of possible values for each of these parameters. We present here the mean of these distributions and their associated 95% credible intervals. The Brooks-Gelman-Rubin (BGR) method was used to monitor convergence of iterative simulations. Convergence was achieved when the upper limit of the credible interval of the BGR diagnostic statistic for a given parameter <1.2 [51].