Help with suspicious estimation results

[jmadeira]

Hi,

When I estimate my model I always get results that are very close to the prior (for most parameters) and with very low standard deviation.
However if I change the prior significantly, I usually obtain again estimates close to the prior and also with low standard errors.

I also obtain very low acceptation rates (close to zero), unless I use a very low jscale (example 0.002). The intervals generated by the MH algorithm are incredibly small.

I find these results suspicious. Does anyone have sugestions on what should I do and why this is happening?

I have in attachment the mode file, the data (some variables are HP filtered), the output results and a pdf file that explains the model

Best,
Joao

[p.gelain]

Hi jmadeira

I think the problem is with your data. I looked at them and they are very volatile. Moreover, almost all the series look like white noise processes. Did you filter them? If yes, maybe you should re-think about the filtering procedure. In the end, the series for "i" and "pi" are quite awful and the series for "m1" display a clear outlier around observation 175. If possible try to remove it, and in general try to have better data. I hope these suggestions will solve your problem.

Best regards.
Paolo

[fabiac]

Dear jmadeira,

I see your post is old, but I'm facing a similar problem. Could you please tell me how you got around this? Thanks