Why the distribution of logsigma can not cross zero?

[Zijian-astro]

I just find that the distribution of logsigma estimated from cont.do_mcmc can not cross zero.
For example, when I run one simulated lc and get a logsigma distribution estimation as this fig:

this look normal, and I expect that when I replace a simulated lc with smaller sigma, the logsigma distribution estimation will globally shift to the left. But in fact I got a fig as follows:

so it seems that the estimation distribution of logsimga can not cross 0, I think this is counterintuitive and I don't why.

Looking forward to your reply. Many thanks.

[nye17]

It's hard to see clearly what the log sigma distribution looks like in the bottom left panel, as the x-axis values are in units of 1e7. Can you zoom in on the plot around the input value of log sigma? Thanks.