Matplotlib put x ticks above bar

Question

Is there any way to put the x-ticks above each stacked bar in a stacked bar chart, rather than below the x-axis? To be clear, I don't mean place the x-tick above each individual bar in a single stacked bar, I mean place the x-tick above the stacked bar itself. Here's how I'm creating my plot:

df = pd.DataFrame(np.random.randint(1, 5, size=(3200, 3)))    
df.loc[np.random.choice(df.index, size=3190, replace=False), :] = 0

df_select = df[df.sum(axis=1)>1]
fig, ax = plt.subplots()

ax.bar(df_select.index, df_select.iloc[:,0], label = df_select.columns[0], wdith = 15)

if df_select.shape[1] > 1:
    for i in range(1, df_select.shape[1]):
        bottom = df_select.iloc[:,np.arange(0,i,1)].sum(axis=1)
        ax.bar(df_select.index, df_select.iloc[:,i], bottom=bottom, label = 
df_select.columns[i], width = 15)

ax.set_xticks(df_select.index)
plt.legend(loc='best', bbox_to_anchor=(1, 0.5))
plt.xticks(rotation=90, fontsize=8) #this puts the x ticks below the x axis

Additionally, I'd like to place some text at specific points along the x axis. I'm storing these sites in a list:

sites = [19, 173, 1002] # the number and elements of this list vary

So, for example, at x = 173, I'd like to place the text 'site (173)' along with a tick at position 173.

For your reference, I've posted images for what my current code produces, and what I would like to produce: current: https://i.sstatic.net/QDlEP.png goal: https://i.sstatic.net/IJJo4.png

Answer 1

This can be achieved by using the example that @Evan linked to in this comments, namely this one. The important bit is the function named autolabel in the linked example. Changing the values displayed from the y axis values to the x axis values is easy, replace height with rect.get_x(). The trickier bit is putting the values at the top of the bar. The total height of your bars can be found by summing the values in your DataFrame.

heights = df_select.iloc[:,:].sum(axis=1)

This then needs to be passed to the autolabel function and used as the height of the bars. The x tick labels can be removed using

ax.get_xaxis().set_ticklabels([])

And you can add further text below your x axis for specific sites by simply using ax.text using the contents of sites as the x location, and setting the y location to be below the axis (-0.5 or something).

Putting this all together, we get the following working example:

def autolabel(rects, heights):
    """
    Attach a text label above each bar displaying its height
    Modified for use with a stacked bar chart
    """
    for i, rect in enumerate(rects):
        x = rect.get_x()
        wid = rect.get_width()
        height = heights.values[i]

        ax.text(x + wid/2., 1.05*height,
                '%d' % (int(x) + int((wid/2)+0.5)),
                ha='center', va='bottom', rotation=90)


df = pd.DataFrame(np.random.randint(1, 5, size=(3200, 3)))    
df.loc[np.random.choice(df.index, size=3190, replace=False), :] = 0

df_select = df[df.sum(axis=1)>1]
fig, ax = plt.subplots()

ax.bar(df_select.index, df_select.iloc[:,0], label = df_select.columns[0], width = 15)

if df_select.shape[1] > 1:
    for i in range(1, df_select.shape[1]):
        bottom = df_select.iloc[:,np.arange(0,i,1)].sum(axis=1)
        rects1 = ax.bar(df_select.index, df_select.iloc[:,i], bottom=bottom, label = 
df_select.columns[i], width = 15)

ax.set_xticks(df_select.index)
ax.get_xaxis().set_ticklabels([])   # turn off the x tick labels

plt.legend(loc='best', bbox_to_anchor=(1, 0.5))

heights = df_select.iloc[:, :].sum(axis=1)
autolabel(rects1, heights)

# Select 3 random value to add labels below the x axis. Make sure they are
# within the axis limits
sites = np.random.choice(df_select.index, 3)

for site in sites:
    ax.text(site, -0.5,
        'site(%s)' % site,
        ha='center', va='bottom',fontsize=6)

plt.show()

Which gives the following graph:

Note: this can sometimes look messy because the bars are very thin and spread out, and can be placed close to each other, meaning values may start to overlap.