Convergence of log Gaussian’s standard deviation

Convergence of partial posteriors’ mean of $sigma$ as a function of sub-sample size for Bardenet’s log-Gaussian model. We randomly sub-sample $n_t$ data from the dataset and run MCMC to estimate $sigma$, whose true value is $sigma=sqrt{2}$. Error bars are 95% over 150 trials.

Leave a Reply

Your email address will not be published. Required fields are marked *