Can you make a self contained complete runnable example that shows the issue? Also report the full trace back? And your matplotlib version (at the very least) and maybe the previous matplotlib version?

Can you make a self contained complete runnable example that shows the issue? Also report the full trace back? And your matplotlib version (at the very least) and maybe the previous matplotlib version?

When the problem occurs, my matplotlib in anaconda is 3.1.0. Then I reinstall the anaconda, and the spyder is 3.3.3, matplotlib is 3.0.3, the .py runs well.

n=9
irf_1 = np.random.uniform(0,1,(100,9))
irf_2 = np.random.uniform(0.5,1.5,(100,9))
irf_3 = np.random.uniform(-0.5,0.5,(100,9))
fig, axes = plt.subplots((n+1) // 2, 2)
for i in range(n):
    irf_data = pd.DataFrame(
            [irf_1[:,i], irf_2[:,i], irf_3[:,i]]).T
    irf_data.columns = ['point', 'up', 'lower']
    
    axes[i//2, i % 2].plot(irf_data.index, irf_data.point,
                           color = 'blue', linestyle = '-')
    axes[i//2, i % 2].plot(irf_data.index, irf_data.up,
                           color = 'red', linestyle = '--')
    axes[i//2, i % 2].plot(irf_data.index, irf_data.lower,
                           color = 'red', linestyle = '--')
plt.show()

This is most probably a duplicate of #14992 and fixed by #15007 (on the matplotlib side) and/or pandas-dev/pandas#27818 (on the pandas side).

What this means is that matplotlib 3.1.1 or lower is not compatible with pandas 0.25.0 when plotting the index of a dataframe.

Options are:

• Use numpy arrays (matplotlib tries to support pandas as good as possible, but there are still some hickups every now and then)
• Wait for matplotlib 3.1.2 to be released.
• Use pandas 0.24.x or 0.25.1 (once it's released)