AttributeError: 'float' object has no attribute '_get_axis_number'

Question

I am trying to get the percentage change in value of today compared to yesterday, for every day in the dataframe. This is the line that throws the error-

import pandas as pd

df = pd.DataFrame({'new_cases':[368060.0,
 357316.0,
 382146.0,
 412431.0,
 414188.0,
 401078.0,
 403405.0,
 366494.0,
 329942.0]})

df['percent_increase_cases'] = df['new_cases'].apply(pd.Series.pct_change)

The formula I am using is

percent_increase = (today's cases - yesterday's cases) / yesterday's cases * 100

It works if I use the code below but I wanted to make it cleaner.

df['percent_increase_cases'] = (df['new_cases'].diff(1)) / df['new_cases'].shift(1) * 100

jjramsey · Accepted Answer · 2021-05-12 13:17:52Z

1

Looks like what is happening is that apply() applies the function pd.Series.pct_change to each element of the series df['new_cases'], rather than to the series as a whole. For example, if I run

pd.Series.pct_change(df['new_cases'])

then I get this:

0         NaN
1   -0.029191
2    0.069490
3    0.079250
4    0.004260
5   -0.031652
6    0.005802
7   -0.091499
8   -0.099734
Name: new_cases, dtype: float64

However, if I apply pd.Series.pct_change to the first element of df['new_cases'], 368060.0, like this,

pd.Series.pct_change(368060.0)

then I reproduce your error:

Traceback (most recent call last):

  File "<ipython-input-7-7aee3ee9524c>", line 1, in <module>
    pd.Series.pct_change(368060.0)

  File "/usr/local/anaconda3/lib/python3.8/site-packages/pandas/core/generic.py", line 10078, in pct_change
    axis = self._get_axis_number(kwargs.pop("axis", self._stat_axis_name))

AttributeError: 'float' object has no attribute '_get_axis_number'

Looks like the solution, then, is to run either

pd.Series.pct_change(df['new_cases'])*100

or equivalently,

df['new_cases'].pct_change()*100

The factor of 100 is there so you get a percent rather than a decimal fraction, consistent with your original formula, (df['new_cases'].diff(1)) / df['new_cases'].shift(1) * 100.

answered May 12, 2021 at 13:17

jjramsey

1,24810 silver badges19 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

callmeanythingyouwant Over a year ago

I can actually swap the series itself for pd.Series! That makes perfect sense! And looks so elegant. Thanks

Gulzar · Accepted Answer · 2021-05-12 15:38:11Z

1

Another, simpler method of doing this would be

df['percent_increase_cases'] = df[['new_cases']].apply(pd.Series.pct_change)

Notice the extra pair of [] when selecting columns.

Selecting a single column from a dataframe returns a series, which would run in to the problems described by @jjramsey, but selecting a list of columns keeps the dataframe as a dataframe, not running into trouble.

answered May 12, 2021 at 15:38

Gulzar

28.8k42 gold badges159 silver badges260 bronze badges

Collectives™ on Stack Overflow

AttributeError: 'float' object has no attribute '_get_axis_number'

2 Answers 2

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related