This is issue 59 in previous repository.
We drafted a test for get_An():
test_that("getAn is non-singular (invertible) for IPW estimation", {
...
})
Notice that this still allows an ill-conditioned (nearly singular) matrix to pass. This can be a low priority issue, as An can be positive semi-definite -- this can occur if the regression models used as Q-functions have collinearity. Still, we may need to enhance this test as a guardrail.
This is issue 59 in previous repository.
We drafted a test for
get_An():Notice that this still allows an ill-conditioned (nearly singular) matrix to pass. This can be a low priority issue, as
Ancan be positive semi-definite -- this can occur if the regression models used as Q-functions have collinearity. Still, we may need to enhance this test as a guardrail.