When estimating models with Bayesian or MLE. The routine in Dynare gives out some value of objective functions. Is a large number or a small number a good sign ?
My hunch is that it should be negative and close to 0, but sometimes the iteration push the result to the opposite direction.
i dont think the value of the likelihood or posterior has any meaning by itself.
in my case, the data is entered in percentages ( ie multiplied by 100), and for the model that i am estimating now, the likelihood value typically starts at -1200 or so…the mode is at -1100. in past experiences, i have noted that if the model is simply a terrible fit on the data, the likelihood value will blow up to say close to -10000000000 and then the optimizer will get stuck there!!!
i have noted that if the value is very large in absolute value, the optimizer stops running and the mode check command that gives us the curvature, gives very spiky looking likelihood graphs.
reuben