Note that I have introduced the mean m(t). Clearly only two of the three of v, a, or m(t) are independent. For a variety of reasons, both operational and historical, it is good to work with m(t), sometimes written just as m, and v.

The mean of the negative binomial distribution is always bigger than the variance (for the Poisson distribution, recall, they are equal). We say then that the data are overdispersed and that v is a measure of the tk a"

0 0

Post a comment