Geweke 1992 convergence test

Estimation ML/Bayesian estimation
1

Hi everyone

I have a question about Geweke (1992) convergence test by Dynare

As the attached file Geweke.pdf (101.7 KB)

Geweke (1992) suggests a convergence test over the null hypothesis of equality of two sample mean
by Chi2-test
I use the p value which is computed based on 15% Taper window of Newey-West standard errors (the last column). If I take the significant level of 5%, then except the 4 following parameters (highlight in the pdf file)

  1. SE_eps_zetachat (p-value = 0.0005)
  2. lambdaf (p-value = 0.027)
  3. lambdami ( p-value = 0.000)
  4. rpie (p-value = 0.001)
    all other parameters in my estimation converge because theirs p-values are greater than the significant level of 5%
    My conclusion is correct. If not, please correct me
    Thank you so much for that
2

Yes, exactly. I would therefore have a look at the trace plots for these parameters. The complication is that the posterior is a joint distribution. Thus, if the null hypothesis of equal means is rejected for one of the parameters, it is unlikely that the full distribution has converged.

3

Dear Prof. @jpfeifer

I got it. Thank you so much indeed for your help