Fix various bugs related to detrending and prefiltering
-
dyn_forecast
adds unlogged steady state if loglinear
is used
-
dynare_estimation_init
sets the noconstant
based on the unlogged steady state if loglinear
is used
- when trends are specified as a function of deep parameters, the values are not correctly updated during estimation due to using the base-workspace parameter values and not the updated local ones
- when using trends with the
prefilter
option, the mean shift due to the trend is not accounted for
- when using
first_obs>1
, the higher trend starting point is not taken into account (leads also to problems in recursive forecasting)