search

StatisticsNZ / demest

Bayesian demographic estimation and forecasting.
Other
21 stars 6 forks source link

Allow use of robust versions of priors when prior is 'saturated' #3

Open johnrbryant opened 7 years ago

johnrbryant commented 7 years ago

A prior is 'saturated' if the associated main effect or interaction has the same length as data 'y'. At present only priors with normally-distributed errors are permitted. To allow t-distributed errors, we need to revisit the updating of 'sigma'.