



access to data and parameters by branch and leaves
--------------------------------------------------

* this could be used in a similar way in other system estimators
* dictionary access to data per equation, branch, leaf
  * alternative dictionary of individual variables used in equations
  * memory: temporary copies while using linalg, or permanent copies
* outsource acces into call-back function or method
* with a parser the dictionaries could be created from a formula
* similar applies for accessing parameters from the params array in optimization (MLE, GMM)
  when there are identical parameters across equations or leaves
  * in latest version use dictionary mapping coefficient/parameter names to index into params

global/outer-scope variables
----------------------------

* for nested logit I need to access params and store, change probabilites from the branch
* need variables in a scope outside of the recursion
* nested function or save params and probs in instance attribute


constraint estimation as alternative and extension
--------------------------------------------------

* cross-equation and within equation restrictions on parameters could also be directly
  imposed, instead of relying on the dictionary
* no idea yet on how to encode this
* might be easier to use constraint optimizer and keep constraints (partially) separate from
  the parameterization in the likelihood calculations

