Bias in resulting estimate ofdistnibution

It is perhaps more surprising that any method of estimating the time of infection by a single time point will result in the estimation of a biased distribution of incubation time. The exact resulting perturbation of the distribution is difficult to predict but plainly, for example, the variance will be overestimated, as the times subject to analysis will have one variance component in themselves (the natural variability of incubation times) and one additional variance component from the estimation of the time of infection. This second source of variance must of necessity be greater than zero, as the true times of infection are unknown. Taylor et al tried to correct for this effect,'7 but we believe that the resulting distribution remains biased. We do not, however, know of any other method of correction.