How long should a dataset be in Bayesian estimation?

In Bayesian estimation, how long should a dataset be? Would 8 years quarterly data be too short?


See this:
Page 6

There is no general answer. Conceptually, Bayesians do not need any data. In this case, your posterior is just your prior. Even one single observation allows you to update prior. But you can never say how much your prior will be updated. Even 1000 observations may be totally uninformative. 8 years might be sufficient if you do not have too many parameters to be estimated. In any case, try it and look at the prior posterior plots.

1 Like

Thank you Jonathanb! This is a really good reference!