Figure 2

Mean-variance relationships for different models. A normal model assumes that the variance is constant and does not depend on the mean. A Poisson model assumes that the variance is equal to the mean. A binomial model has maximal variance at a proportion of 0.5 and progressively smaller variances on either side. The means and variances for the four experimental groups from Mehta et al. are plotted [18], and if the points fall along the line of equality then a Poisson model would be suitable. The variance is greater than the mean in this data, indicating overdispersion. Dashed line = regression line.