| Abstract: | The differential blood count obtained by unbiased sampling mathematically follows a multinomial distribution. The variation between individuals can be formalized by a mixing distribution of the parameters of the multinomial distribution. By DIRICHLET-distributions used as mixing distributions the main phenomena of the observed data can be described, and they are useful in estimating and testing treatment effects. If the correlation between the different types of leucocytes is taken into account in an appropriate manner, univariate test procedures can be applied also. |
