I just finished a whole bunch of tests of the core FHorz features:
Tests with/without d,z,e,semiz
Tests that with/without DC (divide and conquer) gives same answer [and that this is true with/without GI]
Tests that with/without GI (grid interpolation layer) you get same model stats
Tests that lowmemory=1,2 (where appropriate) still give same answer (as lowmemory=0, which is default)
There are also a bunch of ‘cross-tests’ that aim to test that if you ‘disguise an iid as a markov’ then you get the same answer as just using an iid. (and analagously ‘disgusing a markov as semiz’).
Essentially, it is a heap of tests for the ‘standard’ toolkit features in FHorz. And is mostly about testing ValueFnInter_Case1_FHorz() for the ‘standard’ cases (one endogenous state, no behavioural preferences) as well as a bit about StationaryDist_Case1_FHorz(). AllStats and AgeConditionalStats are used, but it doesn’t really test these per se.
All the tests are passed. If anyone wants to see the tests they are all at: