I am setting up a discrete choice agent problem using a value function iteration as an iterative method, rather than EGM (I could not managed to build a DC-EGM for a firm with dicrete and convex adjustment costs)
As a result, I don not have a backward variable, such as Va, to propagate backward : the value function V is enough for the VFI iteration step. My code to set up
ValueError: Some stage taking another non-immediate-successor stage backward [None] as input
I tried to look into the code of the sj package, but could not make sense of the mechanic sitting behind this error. I do not need a backward variable, but it looks like I need to set one up to move one. Am I missing something ? I can provide more elements of the code if needed
I am setting up a discrete choice agent problem using a value function iteration as an iterative method, rather than EGM (I could not managed to build a DC-EGM for a firm with dicrete and convex adjustment costs)
As a result, I don not have a backward variable, such as Va, to propagate backward : the value function V is enough for the VFI iteration step. My code to set up
choice_stage = LogitChoice(value='V', backward=None, index=0, name='inv_choice', taste_shock_scale='taste_shock', f=discrete_cost)Then I set up the continuous chice stage (there are two investment choice available, hence the continous 2D) :
invt_stage = Continuous2D(backward=['V'], policy=['policy_i1', 'policy_i2'], f=iteration, name='invt', hetoutputs=None)Then I get the following error message :
ValueError: Some stage taking another non-immediate-successor stage backward [None] as inputI tried to look into the code of the sj package, but could not make sense of the mechanic sitting behind this error. I do not need a backward variable, but it looks like I need to set one up to move one. Am I missing something ? I can provide more elements of the code if needed
Thank you in advance