ABSTRACT
Applications of parametric statistical distributions in biology may be divided into two broad classes. First, there are those applications in which a specific underlying distributional model is not clearly specified. There is only a vague understanding of the random processes generating the observed data, often coupled with coarse accuracy for the experimental measurements. But there is a second class of applications, albeit in the minority, in which specific assumptions about the real-world processes generating the data may be tenable, and lead to specific distributional models for describing the random variation observed. By concentrating on positive random variables, and the characteristic properties of various distributions for these variables, it is hoped to demonstrate the generality and limitations of the multivariate lognormal distribution for describing random variation. The chapter considers a model in which the vector of growth ratios is a positive random vector.
