Hello everyone
I need clarification on the Markov chain Monte Carlo for panel quantile regression. What are noisy draws? burns? arate?
I used the MCMC option since the other option Nelder-Mead numerical optimization did not give me detailed results. Only coefficients were generated using the Nelder-Mead numerical optimization option. The standard error, pvalues, and confidence intervals were not generated.
I used command : qregpd LogEmissions LogIndu LogGDP LogGDP2 Logmanu LogTrade Apopn, id(Country) fix(Year) quantile(0.7)
Alternatively, I then used the MCMC option as below:
qregpd LogEmissions LogIndu LogGDP LogGDP2 LogManu LogTrade LogAnnPopn, id(Country) fix(Year) optimize(mcmc) noisy draws(1000) burn(500) arate(.5) quantile(0.7)
My results were perfectly generated using the MCMC option but not Nelder-Mead numerical optimization.
Can somebody please explain the differences??
I need clarification on the Markov chain Monte Carlo for panel quantile regression. What are noisy draws? burns? arate?
I used the MCMC option since the other option Nelder-Mead numerical optimization did not give me detailed results. Only coefficients were generated using the Nelder-Mead numerical optimization option. The standard error, pvalues, and confidence intervals were not generated.
I used command : qregpd LogEmissions LogIndu LogGDP LogGDP2 Logmanu LogTrade Apopn, id(Country) fix(Year) quantile(0.7)
Alternatively, I then used the MCMC option as below:
qregpd LogEmissions LogIndu LogGDP LogGDP2 LogManu LogTrade LogAnnPopn, id(Country) fix(Year) optimize(mcmc) noisy draws(1000) burn(500) arate(.5) quantile(0.7)
My results were perfectly generated using the MCMC option but not Nelder-Mead numerical optimization.
Can somebody please explain the differences??