Multivariate statistics for count outcomes are used for continuous outcomes that have naturally skewed distributions. A count outcome is the absolute number of times that something occurs. Something can NOT occur so the "true zero" allows this outcome to be considered continuous. There is a different type of underlying distribution used with this type of continuous outcome. Multivariate statistics for count outcomes allow you to predict for the number of times that something will occur as a count value. Interpretation of the statistics are similar to the methods of logistic regression as odds ratios with 95% confidence intervals are the primary inferences yielded from the analyses and researchers still have to meet statistical assumptions. The choice of multivariate statistical test for a count outcome depends upon the nature of the skewed distribution: Less varied (Poisson regression) or more varied (negative binomial regression).
For Poisson regression, the mean of the count outcome is higher than its associated variance.
For negative binomial regression, the mean of the count outcome is less than its associated variance.
