A method for the analysis of age-stratified antibody prevalence surveys is applied to a previously reported survey of antibody to rubella virus using oral fluid samples in which the sensitivity of the assay used was shown to be compromised. The age-specific distribution of the quantitative results of antibody tests using oral fluids is modelled as a mixture of strong positive, weak positive and negative components. This yields maximum likelihood estimates of the prevalence at each age and demonstrates that, when used in conjunction with mixture modelling techniques, the results of antibody prevalence studies using oral fluids accurately reflect those obtained using sera.