How to modify pandas df in order to adjust starting point of plot for series?

Question

Im trying to plot a dataframe like this:

A = pd.DataFrame([[1, 5, 2, 8, 2], [2, 4, 4, 20, 2], [3, 3, 1, 20, 2], [4, 2, 2, 1, 0], 
              [5, 1, 4, -5, -4], [1, 5, 2, 2, -20], [2, 4, 4, 3, 0], [3, 3, 1, -1, -1], 
              [4, 2, 2, 0, 0], [5, 1, 4, 20, -2]],
             columns=['a', 'b', 'c', 'd', 'e'],
             index=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10])


plt.plot(np.cumsum(A.transpose()))

It looks like this:

However, I would like the first print of the chart to start at 0 for all lines. I tried adding another column according to this, but didn't work. For some reason the index didn't change and kept the newly created column at the end in the plot.

A['s'] = 0
cols = list(A)
cols.insert(0, cols.pop(cols.index('s')))
A = A.loc[:, cols]
plt.plot(np.cumsum(A.transpose()))

You should of course share the code where you try to add the additional point(s), not the one that is working fine. — ImportanceOfBeingErnest, Mar 14 '18 at 12:59
@ImportanceOfBeingErnest don't downvote immediately, give me a chance to improve the question. — hernanavella, Mar 14 '18 at 13:05
I did not downvote and indeed I wrote this comment to give you the chance to improve the question, which you did, so everything is fine, I'd say. — ImportanceOfBeingErnest, Mar 14 '18 at 13:10

score 2 · Accepted Answer · answered Mar 14 '18 at 13:20

Your approach is absolutely correct. The code from the question will produce the desired plot. However, only in matplotlib 2.2. In earlier versions matplotlib would automatically sort the categories alphabetically before plotting, such that row s is last in the axes.

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt

A = pd.DataFrame([[1, 5, 2, 8, 2], [2, 4, 4, 20, 2], [3, 3, 1, 20, 2], [4, 2, 2, 1, 0], 
              [5, 1, 4, -5, -4], [1, 5, 2, 2, -20], [2, 4, 4, 3, 0], [3, 3, 1, -1, -1], 
              [4, 2, 2, 0, 0], [5, 1, 4, 20, -2]],
             columns=['a', 'b', 'c', 'd', 'e'],
             index=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10])

A['s'] = 0
cols = list(A)
cols.insert(0, cols.pop(cols.index('s')))
A = A.loc[:, cols]

plt.plot(np.cumsum(A.transpose()))

plt.show()

In case you cannot use matplotlib 2.2, you may plot the values without labels and set the labels afterwards.

x = np.arange(len(A.columns))
y = np.cumsum(A.transpose()).values
plt.plot(x,y)
plt.xticks(x, A.columns)

score 2 · Answer 2 · answered Mar 14 '18 at 13:22

2

You can use insert to add a new column with all 0's.

A.insert(0, '0', [0]*10)

The first 0 ist the position of your column, in this case the beginning of your dataframe.
'0' is the name of the column. As .plot sorts your columns, you either can use something that comes before your other columns (like probably '0') or you have to reorder your columns in your plot.
[0]*10 are the values of your new column.

answered Mar 14 '18 at 13:22

NK_

361
1
4
11

1

I wonder why naming the column `"0"` is a solution if it is supposed to be named `"s"`? – ImportanceOfBeingErnest Mar 14 '18 at 13:50
@ImportanceOfBeingErnest I skipped this part of his question unintentionally, to be honest. However, I am not sure if he really needs the column to be named 's' or if it is just a dummy name. There are a few possibilities for sorting the columns of a plot. You posted one of them. He can also set the order manually depending on his data. Nevertheless, you're right. – NK_ Mar 14 '18 at 14:26

How to modify pandas df in order to adjust starting point of plot for series?

2 Answers2