Optimal nsample size for Bayesian estimation

[JX2016]

Hi, there. I wonder if there is a rule of thumb for the sample size of the observations used for Bayesian estimation?

Thanks in advance

[JX2016]

If I use the logged 1st-differenced data for the estimation, should I write an equation for dy?

i.e. dy = y - y(-1) and put dy in the varobs block.

Or dynare can understand that y is the 1st-differenced variable so I don't need to specify another variable dy = y - y(-1) ? And I can just treat y as a 1st-differenced variable?


Thanks in advance

[jpfeifer]

1. There is no such thing as the optimal sample size (except for the trivial "as much data as you have from that joint probability distribution"). If your Bayesian, you even do without any data. It is then just called your prior.
2. As detailed in Pfeifer(2013): "A Guide to Specifying Observation Equations for the Estimation of DSGE Models" https://sites.google.com/site/pfeiferecon/Pfeifer_2013_Observation_Equations.pdf the user is responsible for specifying the mapping from model to data variables. That also captures first differences.

[JX2016]

Thank you so much, professor pfeifer! problem solved