用箭头标记matplotlib直方图bin

Question

我有一个直方图,可以用下面的MWE复制:

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

pd.Series(np.random.normal(0, 100, 1000)).plot(kind='hist', bins=50)

这创建了这样的情节:

那么我如何用给定整数的箭头标记bin？

例如,见下文,其中箭头标记包含整数300的bin.

编辑:我应该理想地添加箭头的y坐标应该由它标记的条的高度自动设置 - 如果可能的话!

Answer 1

你可以annotate用来添加一个箭头:

import pandas as pd
import matplotlib.pyplot as plt
#import seaborn as sns
import numpy as np

fig, ax = plt.subplots()
series = pd.Series(np.random.normal(0, 100, 1000))
series.plot(kind='hist', bins=50, ax=ax)
ax.annotate("",
            xy=(300, 5), xycoords='data',
            xytext=(300, 20), textcoords='data',
            arrowprops=dict(arrowstyle="->",
                            connectionstyle="arc3"),
            )

在这个例子中,我添加了一个从坐标(300,20)到(300,5)的箭头.

为了自动将箭头缩放到bin中的值,您可以使用matplotlib hist绘制直方图并获取值,然后使用numpy where查找哪个bin对应于所需位置.

import pandas as pd
import matplotlib.pyplot as plt
#import seaborn as sns
import numpy as np

nbins = 50
labeled_bin = 200

fig, ax = plt.subplots()

series = pd.Series(np.random.normal(0, 100, 1000))

## plot the histogram and return the bin position and values
ybins, xbins, _ = ax.hist(series, bins=nbins)

## find out in which bin belongs the position where you want the label
ind_bin = np.where(xbins >= labeled_bin)[0]
if len(ind_bin) > 0 and ind_bin[0] > 0:
    ## get position and value of the bin
    x_bin = xbins[ind_bin[0]-1]/2. + xbins[ind_bin[0]]/2.
    y_bin = ybins[ind_bin[0]-1]
    ## add the arrow
    ax.annotate("",
                xy=(x_bin, y_bin + 5), xycoords='data',
                xytext=(x_bin, y_bin + 20), textcoords='data',
                arrowprops=dict(arrowstyle="->",
                                connectionstyle="arc3"),
                                )
else:
    print "Labeled bin is outside range"