Creating a an empty dataframe with individual column type set to float

Question

I am creating the columns for an empty dataframe and need to set a particular column (Room #) to float:

columns = ['Time','Name','Company','Room #', 'etc'] # all type object
df["Room #"] = (df["Room #"]).apply(pd.to_numeric, errors='coerce')

However when I check the types at this point, it is int64 instead of the desired float:

df.infer_objects().dtypes
Time                                    object
Name                                    object
Company                                 object
Room #                                   int64
etc                                     object

I don't want to create this as a dictionary because the actual data is fed to the frame later. What is the correct method to coerce the Room # into a float type for that column only?

Answer 1

Use astype:

columns = ['Time','Name','Company','Room #', 'etc'] 
df = pd.DataFrame(columns=columns).astype({'Room #': float})

print(df.dtypes)

Output:

Time        object
Name        object
Company     object
Room #     float64
etc         object
dtype: object

Or, using your method, force the dtype with downcast='float':

columns = ['Time','Name','Company','Room #', 'etc'] 
df = pd.DataFrame(columns=columns)

df['Room #'] = pd.to_numeric(df['Room #'], errors='coerce', downcast='float')

print(df.dtypes)

Output:

Time        object
Name        object
Company     object
Room #     float32
etc         object
dtype: object