Unexpected behavior -- crosstab with dropna=False

[cotterman]

Summary of issue

By the name of it, I expected using the 'dropna=False' option with crosstab would have given me results that include rows for which my column value is missing. However, I instead found that missing values continue to be ignored.

Code Sample

 df = pd.DataFrame({'geography': ['tropic', 'tropic', 'desert', None, 'tundra', 'tundra'], 
                    'dragons': ['many', None, 'few', 'few', 'many', None]})
 print(df)
 pd.crosstab(df.geography, df.dragons)
 pd.crosstab(df.geography, df.dragons, dropna=False)
 pd.crosstab(df.dragons, df.geography, dropna=False)

Expected Output

geography  NA  desert  tropic  tundra
dragons                              
NA          0       0       1       1
few         1       1       0       0
many        0       0       1       1

Actual output

In [210]:  print(df)
  dragons geography
0    many    tropic
1    None    tropic
2     few    desert
3     few      None
4    many    tundra
5    None    tundra

In [211]:  pd.crosstab(df.geography, df.dragons)
Out[211]: 
dragons    few  many
geography           
desert       1     0
tropic       0     1
tundra       0     1

In [212]:  pd.crosstab(df.geography, df.dragons, dropna=False)
Out[212]: 
dragons    few  many
geography           
desert       1     0
tropic       0     1
tundra       0     1

In [213]:  pd.crosstab(df.dragons, df.geography, dropna=False)
Out[213]: 
geography  desert  tropic  tundra
dragons                          
few             1       0       0
many            0       1       1

This work-around gives me what I expected.

pd.crosstab(df.dragons.fillna('NA'), df.geography.fillna('NA'), dropna=False)

I'd hate for others to also not realize that the NAs are being dropped from the output, even when dropna=False. Perhaps altering the documentation or the behavior of this function would help.

[sinhrks]

Thanks for the report. Dupe with #10772.