Non covariate variables being detected as covariates

[tomicapretto]

Thanks to #796 I found that the get_model_covariates() function is returning non-covariate variables as covariates.

Example:

import bambi as bmb
import numpy as np
import pandas as pd

from bambi.interpret.utils import get_model_covariates

df = pd.read_csv("https://raw.githubusercontent.com/m-clark/generalized-additive-models/master/data/pisasci2006.csv")

knots_income = np.linspace(np.min(df['Income']), np.max(df['Income']), 5+2)[1:-1]
formula = "Overall ~ 1 + bs(Income, degree=3, knots=knots_income)"
model = bmb.Model(formula, df, dropna=True)
get_model_covariates(model)

array(['Income', 'knots_income'], dtype='<U12')

The non covariate variable is knots_income. The variable is not looked up into the data frame. It comes from the outer environment.

[GStechschulte]

Hey @tomicapretto I was going to assign this to myself, but realized you already have a hotfix branch to resolve this. Do you want me to take this over, or are you fine completing it?

P.S. You indeed found the correct issue 😄

[tomicapretto]

Ha, thanks! I'm completing it right now ;)