Understanding forecasts

I make forecasts as part of the estimation command. Unfortunately, I’m unsure how to interpret the output, in particular the “red lines”, HPDTotalInf, HPDTotalsup. These supposedly include both parameter and shock uncertainty. With my estimation they mainly look spurious: The interval between HPDTotalInf & HPDTotalsup doesn’t include the interval between HPDInf & HPDsup (green lines) as a subset. For several variabels and horizons the two intervals don’t even overlap at all. Looking at the code of forecast.m , forcst.m and simultexdet.m didn’t help for me. In my uncertain judgement the variabels in question seem to be generated by the line:
"[1 k], repmat(dr.ys(ivar(i), 1, 2), ‘-r’)" in forcst.m.

Do I totally misunderstand the definition of HPDTotalInf, HPDTotalsup (or alternatively the def. of shock uncertainty). Is there maybe a setting for how many shocks to draw for this computation?

I faced the same problem. The intervals with the total uncertainty didn’t behave properly (more or less in the same way you describe). After a first look at the functions, I suspected on the “simul_seed” option. Though I’m not 100% sure on this, I believe there is a loop that re-initializes the random number generator at the same seed when drawing the simulated shocks.

I have attached two figures with the forecast for the same model (a simple RBC). I run the model twice. All parameters and options are identical except for the inclusion of the simul_seed option (=10) in the estimation sentence in one of them.

Can someone confirm me whether this is a small bug?

RBC_model_Bay_Forecasts1.pdf (6.9 KB)
Copy_of_RBC_model_Bay_Forecasts1.pdf (7 KB)