multiple chains and posterior plot

[DW916]

Dear Johannes,

There has been a question confusing me for a long time. Hope seek some help from you.

If I run 3 chains in estimation, each chain 100,000 draws, drop first half. Then how to explain the normal distributed posterior plot? Does that mean total 50,000+50,000+50,000 draws distribution? How can I know each chain's draws?

Many thanks.

Catherine

[jpfeifer]

The prior posterior plots pool the draws from the the different chains after the burnin, use a subsample from these draws and then use a kernel density estimate.

There is currently no way to select draws from a particular chain. You can only use

Code: Select all

trace_plot

to plot draws from the different chains.