In a similar vein to the note I made on issue #89, cov_adj needs to handle a model whose design matrix and fit are less than full rank. As I did in the other issue, I will adopt the stats package's method of handling it: when making predictions, drop the columns of the design matrix whose corresponding regression coefficients weren't estimated.
In a similar vein to the note I made on issue #89,
cov_adjneeds to handle a model whose design matrix and fit are less than full rank. As I did in the other issue, I will adopt thestatspackage's method of handling it: when making predictions, drop the columns of the design matrix whose corresponding regression coefficients weren't estimated.