Dynare forums

is it possible to truncate the prior in a bayesian estimation exercise in dynare?
say if you have a normal prior, can you truncate it to cover only the non-negative domain?

reuben

when you declare the prior, you can specify bounds for the parameter. it's done by adding the bounds after the mean and std.dev of the prior
ex:sigma, inv_gamma_pdf, 2, .2, 0.01, 3;
the last two numbers specify the interval you allow the parameter to be in.
Cosmin

thanks. does the truncation guarantee that the prior density will integrate to unity? if you truncate it, the constant of integration would change, right.


reuben

You're right. I am not sure, but the integration to unity should happen by changing the constant.
Cosmin