mode_compute=4 and measurement error

[sarah]

Hi again, I am doing a bayesian estimation. Everythink works fine (thanks to your kind suggestions).

Now I am trying to consider a measurement error that I declare directly in the bayesian estimation as suggested by Pfeiferhttps://sites.google.com/site/pfeiferecon/Pfeifer_2013_Observation_Equations.pdf?attredirects=0.

Code: Select all

stderr x_ob, uniform_pdf,,,0,0.1; //measurement error

Unfortunately I am not able to use mode_compute=4 with the m_error. I tried with 6 but I end up with bad estimation.


Do you know if there is some kind of incompatibility between measurement error and mode_compute=4? Have someone succeeded in using compute=4 with measurement error declared in this way? Should I change the way in which I declare m_error? Thanks!

[jpfeifer]

Hi, what do you mean with "bad estimation"? mode_compute=4 in general does not always perform well with bigger models. You might need to use a different combination of mode-finders. Note also that you will run into trouble if the measurement error estimate hits the upper bound as this will be a corner solution with a non-positive definite Hessian. In that case, you might want to use a different, more informative prior.