Great article! I find very useful to combine some causality approaches with more common machine learning models we use in our daily tasks in Data Science. One question though: where did the df_norm come from? You started with df and later presented a dataset named df_norm.