Getting error "TypeError: no numeric data to plot" in a Time Series ARIMA analysis

Question

I am trying to follow a tutorial whereby an ARIMA time series analysis using differenced data is being done:

The following is the python code:

def difference(dataset):
    diff = list()
    for i in range(1, len(dataset)):
        value = dataset[i] - dataset[i - 1]
        diff.append(value)
    return Series(diff)

series = pd.read_csv('dataset.csv')
X = series.values  # The error in building the list can be seen here
X = X.astype('float32')
stationary = difference(X)
stationary.index = series.index[1:]
...
stationary.plot()
pyplot.show()

When the process reaches the plotting stage I get the error:

TypeError: no numeric data to plot

Tracing back, I find that the data that is being parsed is resulting in a collection of array. Saving the collection stationary as *.csv file gives me a list like:

[11.]
[0.]
[16.]
[45.]
[27.]
[-141.]
[46.]

Can somebody tell me what is going wrong here?

PS. I have exluded the parts of import of libraries

Edit 1

A section of the dataset is reproduced below:

Year,Obs
1994,21
1995,62
1996,56
1997,29
1998,38
1999,201

Answer 1

To difference, just use Series.diff or DataFrame.diff().

Also, Year should be the index:

from matplotlib.ticker import MaxNLocator
import numpy as np
import pandas as pd

df = pd.DataFrame(
    {'Year': np.arange(1994, 2000),
     'Obs': [21, 62, 56, 29, 38, 201]})

stationary = df.set_index('Year').diff()
ax = stationary.plot(legend=False)
ax.xaxis.set_major_locator(MaxNLocator(integer=True))

Output:

Answer 2

pd.read_csv('dataset.csv') returns a pandas.DataFrame, not a pandas.Series.

You can read the data as a pandas.DataFrame, then select the columns you want in each axis. The selected column are pandas.Series. Column Year is converted to str so it doesn't show as float.

df = pd.read_csv("dataset.csv")
y = difference(df["Obs"])
x = df["Year"].astype(str)[1:]
pyplot.plot(x, y)
pyplot.show()

This will plot: