CODEWITHSUNDEEP
pythonpandas
Gabriel
Gabriel

Reputation: 3797

mean calculation in pandas excluding zeros

Is there a direct way to calculate the mean of a dataframe column in pandas but not taking into account data that has zero as a value? Like a parameter inside the .mean() function? Was currently doing it like this: 

x = df[df[A]!=0]
x.mean()

Upvotes: 41

Views: 81276

Answers (4)

nuri
nuri

Reputation: 240

df[df["Column_name"] != 0]["Column_name"].mean()

or if your column name does not contain space char

df[df.Column_Name != 0].Column_Name.mean()

hopefully it can be included as a parameter in the next "mean" object version

.mean(exclude=0)  #wondering in next versions

Upvotes: 2

SimonM68
SimonM68

Reputation: 21

Very late to the discussion, but you can also do:

df[df["Column_name"] != 0].mean()

Upvotes: 0

Александр Ласкорунский
Александр Ласкорунский

Reputation: 21

You can convert df to numpy array and use numpy.nanmean()

import numpy as np

df = pd.DataFrame(data=np.array([[1, 2], 
                                 [3, 4], 
                                 [6, 7], 
                                 [8, np.nan], 
                                 [np.nan, 11]]), 
                  columns=['A', 'B'])

df_col_means = numpy.nanmean(df.values)  # by columns
df_row_means = numpy.nanmean(df.values, axis=1)  # by rows
col_A_mean = numpy.nanmean(df['A'].values)  # particular column mean

Upvotes: 0

tibi3000
tibi3000

Reputation: 618

It also depends on the meaning of 0 in your data.

  • If these are indeed '0' values, then your approach is good

  • If '0' is a placeholder for a value that was not measured (i.e. 'NaN'), then it might make more sense to replace all '0' occurrences with 'NaN' first. Calculation of the mean then by default exclude NaN values.

    df = pd.DataFrame([1, 0, 2, 3, 0], columns=['a'])
df = df.replace(0, np.NaN)
df.mean()

Upvotes: 49

Related Questions