You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
we need constraints or contrast matrices in several places,
original definition is in t_test and similar.
test_wald_terms uses some string matching to construct constraints
fit_constrained uses the same pattern, IIRC
penalized fit needs restriction matrices to define stochastic constraints
proposal:
I think we can create a standalone function following the t_test pattern.
It would be convenient to allow for wild cards as in Stata,
e.g. "region*" for all parameters that encode a categorical variable
The latter could by also used with sparse matrices, e.g. as in variance components in mixed models, or variation on sparse OLS and sparse penalized models.
get_constraints(constraints_string, param_names)
or some similar name
possible extension to automatic constraint patterns
e.g. all pairwise #4365
The text was updated successfully, but these errors were encountered:
a general idea
see also related, more general issue #4361
we need constraints or contrast matrices in several places,
proposal:
e.g. "region*" for all parameters that encode a categorical variable
get_constraints(constraints_string, param_names)
or some similar name
possible extension to automatic constraint patterns
e.g. all pairwise #4365
The text was updated successfully, but these errors were encountered: