Dear all, I am trying to replicate Christensen, Dib (2008)'s estimation results using "EDSGE" function in BMR package. Attached are my R code, the data I collected and preprocessed, and the paper. I set the priors following the literature (e.g., Smets and Wouters, 2007). However, when I ran the code, R reported error messages like: Beginning MCMC run, Mon Jan 23 22:42:03 2017. Error in chol.default(CovM) : the leading minor of order 2 is not positive definite In addition: Warning message: In sqrt(parModeHessian) : NaNs produced I also found that the estimation is sensitive to initial values and/or prior settings. I am struggling with debugging the code and hope someone could help me out. Thanks! Best, April -------------- next part -------------- A non-text attachment was scrubbed... Name: The financial accelerator in an estimated New Keynesian model_Christensen, Dib (2008 RED).pdf Type: application/pdf Size: 2680535 bytes Desc: The financial accelerator in an estimated New Keynesian model_Christensen, Dib (2008 RED).pdf URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20170123/1d3e2958/attachment-0001.pdf>
Replicate Christensen, Dib (2008) using BMR package
2 messages · Xue Li, Jeff Newmiller
Only your pdf attachment made it through. You would need to follow the Posting Guide regarding file types to achieve successful transmission. I will say that open-ended debugging of your code does not fit the Posting Guide recommendations either... this is an R language mailing list, not a research assistance mailing list. It might fit in better at stats.stackexchange.com, but even they want a focused question rather than general assistance requests.
Sent from my phone. Please excuse my brevity. On January 23, 2017 7:09:30 AM PST, Xue Li <xuelimizzou at hotmail.com> wrote: >Dear all, > > >I am trying to replicate Christensen, Dib (2008)'s estimation results >using "EDSGE" function in BMR package. Attached are my R code, the data >I collected and preprocessed, and the paper. > > >I set the priors following the literature (e.g., Smets and Wouters, >2007). However, when I ran the code, R reported error messages like: > > >Beginning MCMC run, Mon Jan 23 22:42:03 2017. >Error in chol.default(CovM) : > the leading minor of order 2 is not positive definite >In addition: Warning message: >In sqrt(parModeHessian) : NaNs produced > >I also found that the estimation is sensitive to initial values and/or >prior settings. I am struggling with debugging the code and hope >someone could help me out. Thanks! > > >Best, > >April