This major release changes the version support for several items. In order to speed up the computation of the bread matrix, SciPy approx_fprime funtionality is now used. This required the following version changes:
- SciPy v1.9.0+ which requires NumPy 1.18.5. This is a major change in the supported versions - SciPy v1.9.0 also necessitates the reduction of support to 3.8+
The regression estimating equations have also now been fully updated to the new syntax. The legacy versions have been removed.
0.6
Additional robust loss functions, better errors when nan's appear in the estimating equations, added warnings for non-differentiable estimating equations
0.5.0
v0.5 adds the optional `subset` parameter which allows a user to optimize only a subset of the parameters (during the root-finding phase). This is useful when combined with pre-washing steps (pre-washing is essential for proper use of the `subset` functionality).
* IGNORE the 0.5.0 tag. This is a GitHub issue (it things v0.5 is an existing tag, which it is not)
0.4
v0.4 adds penalized regression estimating equations, updated the syntax structure for regression equations (with the previous API being dropped in v1.0), and updated the internal structure for built-in estimating equations for easier maintenance.