Move X-Axis label downwards, but not X-Axis Ticks in matplotlib

Question

I'm using Matplotlib to plot a histogram. Using tips from my previous question: Matplotlib - label each bin, I've more or less got the kinks worked out.

There's one final issue - previously - the x-axis label ("Time (in milliseconds)") was being rendered underneath the x-axis tickmarks (0.00, 0.04, 0.08, 0.12 etc.)

Using the advice from Joe Kingston (see question above), I tried using:

ax.tick_params(axis='x', pad=30)

However, this moves both the x-axis tickmarks (0.00, 0.04, 0.08, 0.12 etc.), as well as the x-axis label ("Time (in milliseconds)"):

Is there any way to move only the x-axis label to underneath the three rows of figures?

NB: You may need to open the PNGs below directly - Right Click on the image, then View Image (in FF), or Open image in new tab (Chrome). The image resize done by SO has rendered them nigh unreadable

Kea

Answer 1

use labelpad parameter:

pl.xlabel("...", labelpad=20)

or set it after:

ax.xaxis.labelpad = 20

Gibbs

if I use this and save the figure, the figure is cropped and I don't see the label anymore – Saintpierre

Answer 2

If the variable ax.xaxis._autolabelpos = True, matplotlib sets the label position in function _update_label_position in axis.py according to (some excerpts):

    bboxes, bboxes2 = self._get_tick_bboxes(ticks_to_draw, renderer)
    bbox = mtransforms.Bbox.union(bboxes)
    bottom = bbox.y0
    x, y = self.label.get_position()
    self.label.set_position((x, bottom - self.labelpad * self.figure.dpi / 72.0))

You can set the label position independently of the ticks by using:

    ax.xaxis.set_label_coords(x0, y0)

that sets _autolabelpos to False or as mentioned above by changing the labelpad parameter.

Welfarism

For posterity, matplotlib now has .set_label_coords which gives a similar level of control – Qianaqibla

Answer 3

Instead of padding manually, try assigning vertical alignment (va=):

ax.set_xlabel("x label", va='top');

If you modify the label padding manually as in @HYRY's answer, sometimes when you save the figure the label is cropped off. To include the label, set bbox_inches='tight' when you call savefig(). For example:

plt.plot(range(10))
plt.xlabel("x label", labelpad=30);
plt.savefig('img.png', bbox_inches='tight');

With the OOP API, the same would be:

fig, ax = plt.subplots()
ax.plot(range(10))
ax.set_xlabel("x label", labelpad=30);
fig.savefig('img.png', bbox_inches='tight');

Linzer