CODEWITHSUNDEEP
pythonnumpypandas
Sanchit Aluna
Sanchit Aluna

Reputation: 465

Making a new variable numerical column using if/else statements

I have a dataset that contains a column 'y' in which a particular values exist. I would like to take that column and make a new column (z) denoting if y value is 47472 then z should be 1000, if y <1000 then z=y*2, else all other values should be 2000. Here's a mock example of the data. I don't have a 'z' column, but I want to create it:

          y      z
0      1751   2000
1       800   1600
2     10000   2000
3       350    700
4       750   1500
5      1750   3500
6     30000   2000
7     47472   1000


def test(y):
    if y == 47472:
        z=1000
    elif y < 1000:
        z=y*2
    else:
        z=2000
    return Z

# I tried to call the above function below
z = test(y)
z

but I don't get the result instead it shows below error:

ValueError: The truth value of a Series is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().

Upvotes: 2

Views: 6276

Answers (1)

agold
agold

Reputation: 6276

The problem is that you are using a Series in the if statement, such as:

if y == 47472:

assuming that y is part of your DataFrame this will result in a list of booleans:

>>> df['y']==47472
0    False
1    False
2    False
3    False
4    False
5    False
6    False
7     True
Name: y, dtype: bool

Which is not legal, and therefore it suggests you to use a boolean function that returns one boolean such as any(), all(), etc. Instead you should use boolean indexing:

# df is the dataframe with your data
# adding column z
df['z'] = pd.Series(np.zeros(df.shape[0]))
# if y == 47472 then put 1000
df.loc[df['y']==47472, 'z'] = 1000
# filter <1000
df.loc[df['y']<1000, 'z'] = 2*df['y']
# now set rest to 2000 (i.e. ones that do not comply previous 2 conditions)
df.loc[(df['y']>=1000) & (df['y']!=47472),'z'] = 2000

Edit: As commented by EdChum I was performing chained indexing:

df['z'][df['y']<1000] = 2*df['y']

which should be avoided by using loc:

df.loc[df['y']<1000, 'z'] = 2*df['y']

Upvotes: 1

Related Questions