+ Fix coefficients so that shape is (n_coefficients,) instead of (n_coefficients, 1) + Add a parametric bootstrap function + If the weighted least squares SVD fails, make all the coefficients NaN.
0.0.4.dev0
+ Handle case when response is all zeros + Make non-independent columns NaN instead of excluding them + Make response always 2D
0.0.3.dev0
Some minor changes: + results are returned as a named tuple so they are easier to understand + renamed some variables for clarity + Filled in some docstrings
0.0.2.dev0
+ Fixed some issues with getting the right linearly independent columns for the summary statistics + Added more summary statistics + Fixed the log likelihood and deviance weights