Consider the toolkit function that calculates the agents distribution for finite horizon models with fixed types:
https://github.com/vfitoolkit/VFIToolkit-matlab/blob/master/StationaryDist/FHorz/PType/StationaryDist_Case1_FHorz_PType
I don’t understand the following comment (in the header comments of the function):
“jequaloneDist can either be same for all permanent types, or must be passed as a structure”
In my model the initial distribution at age 1 depends on permanent types and I define it in the code as an array with size [n_a,n_semiz,N_i], since the state variables are (a,semiz) and there are N_i fixed types in total. All individuals start with zero assets but the initial distribution of semiz varies across types: low skilled agents may have different distribution of semiz compared to high skilled agents, for example.
Does this mean that I have been doing things in the wrong way? 
Comment is out of date. That used to be the case, but now you have essentially three options. Pretend the model state space is just [n_a,n_z].
i) Input jequaloneDist as [n_a,n_z], this will be interpreted as applying to all N_i permanent types
ii) Input jequaloneDist as [n_a,n_z,N_i], this will be interpreted as that the last dimension indexes the N_i permanent types.
iii) Input jequaloneDist as a structure, the fieldnames should match Names_i, and each one of the fields should be a matrix of [n_a,n_z]
Note that if you have a model where the state space differs by permanent type, obviously only the third of these is usable.
Note: If N_i=prod(n_z), the code will get confused. Avoid this.
I’ve cleaned up the comment in code to reflect this.
2 Likes
Thanks for your answer. Now I feel better 
1 Like
