Estimation problems: mode_check_plot and bayesian irf

[Haydenknows]

Hi professor, I came across some questions when estimating the parameters.

1. When the mode_check_plot reports some red dots, may I just ignore the those red dots? Does that mean that the posterior distribution or smoothed decomposition has some errors when those red dots shows up?
   

   

2. The HPD intervals of bayesian irf is almost zero. Is that feasible? How can I solve this problem?
   

   

[jpfeifer]

See my reply at Estimation weird mode_check plot and posterior - #10 by jpfeifer
The problem is that you run into a bound and the jumping proposal becomes tiny to not violate the bound (tiny mh_jscale)