I’m trying to create a new variable in a dataframe that requires some complex conditional logic. I’d like the default value to be np.nan if all of the conditions fail.

my approach has been like this:

startingdf['name_of_new_variable'] = pd.Series(pd.nan, index = startingdf.index).case_when(
  caselist=[
     (...,...),
     (...,...),
     (...,...),
     #etc etc
  ]
)

The key thing I’m hoping for here is that case_when will be called on a series populated with pd.nan (hence pd.Series(pd.nan...)). That way there will be an automatic pd.nan in any row in the name_of_new_variable column where none of the conditions in the caselist are met.

But there’s something wrong with pd.Series(pd.nan, index=startingdf). When I write…

pd.Series(pd.nan, index = startingdf.index).case_when()

my typechecker (happens to be Pylance) complains that “object of type float is not callable.”

So somehow, pd.Series(pd.nan, index=startingdf.index).case_when is returning a float instead of the (callable!) method case_when of a Series instance!

What am I doing wrong?