It looks like the design of the GLM function patsy takes the predictors as column names of pandas dataframe. So wondering if there a way to declare the predictors as theano shared variables when using the PyMC3 GLM for performing logistic regression so that you can use the shared variables to define test data. If there is indeed a way, can someone please point me to a code snippe? Thanks in advance
I don’t use the GLM model a lot, but in I think you can use the
pm.Data container, as with classical models.
Hope this helps