Statistical models predicting the prevalence of light, moderate and heavy intensity infections from total prevalence. Coefficients describe the relationship between the transformation of total prevalence f(x) and log of the relative risk between moderate or heavy infection and light infection ln[π(yi=k|xi)/π(yi=r|xi) (see equation 1 for details). Negative intercepts indicate that when total prevalence is zero, log of the risk ratio is negative. In other words the prevalence of light intensity infection is expected to be higher than the prevalence of moderate or heavy infection. Furthermore, please note that when total prevalence is 0, predicted prevalence is also 0 for all three infection intensities regardless of the intercept. This is achieved by regressing prevalence conditional on infection, then multiplying it by total prevalence to convert it to prevalence among the whole population. Coefficients for light infection are not included as light infection prevalence is derived from moderate and heavy regressions (see equations 2-3). se = standard error of coefficients; qlogis (total prevalence) = logistic transformation of total prevalence