If the variance for the model is multiplicative, then the
calculations change slightly. The same prior applies, but the
posterior for 
may be calculated as follows, where the
number of timepoints observed is n ;
and noting that in this case
The posterior will also be inverse gamma, since
where
