Seasonality is considered in the conversion for the annual frequency to monthly frequency.


Seasonality is implemented by a means of "normalized to 12" in the Simulator as followed:

To illustrate how seasonality affects the conversion of annual frequency, let’s assume that the user inputs a Poisson “annual frequency” with lambda= 4800, and by applying trend, exposure, etc, other factors, we already have "monthly frequency" with lambda= 400. T he seasonality vector is <2,1,1,2,1,1,2,1,1,2,1,1>.

  1. The total of the vector elements above is 16. then thenormalize ratio is 16/12= 1.333333
  2. For months 1, 4, 7, and 10,the seasonality is then normalized to 2/1.33333= 1.5
  3. For other months, the normalized seasonality is 1/1.3333333 = 0.75
  4. Then the "monthly frequency" parameter lambda for months 1, 4, 7, and 10 is 400 x 1.5 = 600, and the parameter for the other months would be 400 x 0.75 = 300.


The entire discussion above carries through for the negative binomial distribution also.  For this distribution, the “size” parameter would be treated in a manner analogous to lambda   for the Poisson distribution.