Dynare forums

Hi all,

How to impose inequality constraints in Dynare? I have to estimate say, apha1 and alpha2, with alpha1 and alpha2 positive and alpha1 > alpha2. If the model is estimated with Bayesian methods, is it sufficient to specify such constraints when defining the prior.
For instance, choosing a gamma for alpha1 and alpha2 with different mean. Correct?

Thank you,
F.

The simplest solution is to choose priors for alpha1 and alpha2 with a bounded support and sufficiently tight to ensure that the two distributions do not overlap.

Otherwise, I suggest to redefine the parameters: create a parameter (say "p") such that alpha1 = alpha2 + p, and enforcing the positivity of p to have a positive support.

Of course this means that in order to compute the posterior distribution of alpha2 you will have to do some computation.

Note that I have never tried this solution and that I am not even sure that it works in practice. You may want to ask confirmation from a Bayesian expert.