Page 1 of 1

Bayesian model averaging

Posted: Thu Apr 05, 2012 2:45 pm
by EViews Esther
This thread is about the BMA add-in which accounts for model uncertainty associated with variable selection.

Re: Bayesian model averaging

Posted: Sun Aug 25, 2013 10:43 pm
by ESoliman
Hi, I'm trying to use the BMA EViews add-in for model averaging (already used the R counterpart package successfully). Also went through some pages of the Command Reference (pp. 154 onwards) on the requirements to properly run the BMA add-in. I think I've checked all necessities already (rscproxy, statconnDCOM, and R). Couldn't 'sync' it though, for example when I tried the StatConnector Test the error "Method ~ of object '~' failed" kept on showing itself.

Also tried to use a version of RandFriends that I found for an easier install. I thought the bma() EViews function would already work but after a little waiting time a different error popped-up ("R (for bma<-bic.glm(x,...) returned an error; evaluation stopped because of an error"). Could you give some specific recommendations for BMA to work properly in EViews? Thank you.

Re: Bayesian model averaging

Posted: Sun Aug 25, 2013 11:14 pm
by EViews Gareth
Sounds like you got it working, it just doesn't work on your particular data.

Re: Bayesian model averaging

Posted: Wed Aug 28, 2013 12:40 am
by ESoliman
Hi Gareth. I tried a different EViews workfile, a simple one with 74 time series observations with 15 variables. Utilized the needed commands (e.g. xopen()) to maintain the connection to make sure. My target was to regress a response variable on the rest of the series. The required window appeared after clicking "Add-ins" >>> "BMA for GLM/LM". However after writing the series names on the pertinent box and pressing "OK", the error "R (for 'temp_EViews_depvar') returned an error; internal error: unsupported data format passed to interpreter." popped up.

Tried it using the "bma(method=LM ..." technique as well but I'm getting the same error. Can you advise on a possible workaround to this issue? Is it still possible to execute bayesian model averaging using the BMA add-in despite this?

Thanks,
Edward

Re: Bayesian model averaging

Posted: Thu Sep 12, 2013 10:41 am
by Jamel Saadaoui
Dear Sir,

I have some trouble when I try to use the BMA add-in with Eviews.

When I try to run a regression, Eviews send me an error message "StatConnectorSrv.StatConnector is not connected".

I downloaded the latest version of R and of the BMA package.

Thank you for your help,

Jamel

Re: Bayesian model averaging

Posted: Thu Sep 12, 2013 11:01 am
by EViews Gareth
You need to get statconnector working properly. Follow the instructions for installing RExcel here:
http://homepage.univie.ac.at/erich.neuw ... to_install

Re: Bayesian model averaging

Posted: Thu Sep 12, 2013 11:39 am
by Jamel Saadaoui
Thank you very much for your help, now it's work perfectly with some data downloaded thanks to getmacrodata add-in.

Jamel

Re: Bayesian model averaging

Posted: Tue Oct 14, 2014 12:33 am
by uwhusky
Hi,

When I tried to run the BMA add-in in EViews 8, EViews prompts me the error message "No external connection is open." Any clues why this message is shown, and how to resolve this problem? Thanks.

Re: Bayesian model averaging

Posted: Tue Oct 14, 2014 8:17 am
by EViews Gareth
Do you have R properly installed and working?

Re: Bayesian model averaging

Posted: Thu Mar 09, 2017 12:57 pm
by Geo
Hi everybody (I hope that someone will see this post!)

When I am trying to run BMA the following message appears: "StatConnectorSrv.StatConnector is not registered".

I have downloaded all the necessary libraries in R.

I tried to use the linked suggested before (http://homepage.univie.ac.at/erich.neuw ... to_install),
but I got totally lost!

Any ideas?
Thanks.

Re: Bayesian model averaging

Posted: Tue Mar 10, 2020 8:06 am
by nlevenko
Hi, Gareth,
The link you gave is not active anymore. Could you give a valid link please?
Thanks!

Re: Bayesian model averaging

Posted: Tue Mar 10, 2020 8:36 am
by EViews Gareth
statconnector is no longer required.