Why does this code yield a NaN in my pandas dataframe?

Question

I'm trying to write some code to predict outcomes of a sports season. I have a dataframe with an int32 Year column, an object Team column for listing the names of the teams, an int32 Capacity column with stadium capacities, and an int32 Attend/G column with average attendance per game played in a given year. There are no previous NaNs within the data.

This line of code perfectly overwrites the stadium capacity for team NYM before a given year

train.loc[(train.Year < 2009) & (train.Team == 'NYM'), 'Capacity'] = 57333

However, the code shown below fails to do so, as it converts the capacity for team PHI to a NaN, and the capacities for the other teams to floats.

I assigned a variable to store the highest Attend/G number for a team. The variable is an int32.

max_attend_per_g_phi = (train.loc[train['Team'] =='PHI',['Attend/G']].max())
print(max_attend_per_g_phi)
print(train.info()
train.loc[(train.Year < 2012) & (train.Team=='PHI'),'Capacity']=max_attend_per_g_phi
print(train.info())

The first train.info call outputs

Attend/G        360 non-null int32
Capacity        360 non-null int32
Team            360 non-null object
Year            360 non-null int32

While the second call outputs

Attend/G        360 non-null int32
Capacity        355 non-null float64
Team            360 non-null object
Year            360 non-null int32

Overwriting the stadium capacity with that variable in the same format as the first line of code yields a NaN for the PHI capacity, as well as converts the other capacities in the dataframe from ints to floats.

Does anyone know why this is occurring?

Answer 1

In some cases working with Pandas for instance when read from CSV pd.read_csv(na_values = "?") you may check your data afterwords like this:

dataset.isna().sum()

This may bring the feedback if your data has the NaN inside the original document.