“pandas crosstab” margin NaN behavior

Question

I am confused about the marginal subtotals of Pandas crosstabs.

For example, let df1 and df2 be Pandas DataFrames:

df1 = pd.DataFrame({1: [2], \'foo\':