如何使用 matplotlib 子圖進行多行布局

Question

這是一個關於如何正確組織子圖的問題，而不是如何創建堆疊條。

我有以下數據框：

     corpus group  mono p  non p  plus p  minus p
0  fairview   all      49     51      49        0
1      i2b2   all      46     54      46        0
2    mipacq   all      44     56      43        1

並希望按照兩個附圖中給出的輸出進行排列，以便我得到 ncolumns 和 2 行，而不是兩個單獨的子圖，每個子圖各有 1 行（因此在這種情況下，將有 2 行 3 列）單個子圖而不是 2 個子圖上的 1 行 3 列）：

我使用以下代碼將這兩個圖作為單獨的子圖生成：

data = <above dataframe>
semgroups = ['all']
corpus = ['fairview', 'i2b2', 'mipacq']
for sg in semgroups:

    i = semgroups.index(sg)
    ix = i + 7

    ncols = len(set(data.corpus.tolist()))
    nrows = len(set(data.group.tolist()))

    fig = plt.figure()
    
    fig, axs = plt.subplots(1, ncols, sharey=True)

    for ax,(idx,row) in zip(axs.flat, data.iterrows()):
        # I WANT TO PLOT BOTH ROWS on same subplot
        #row[['mono p', 'non p']].plot.bar(ax=ax, color=['C0','C1'])
        row[['plus p', 'minus p']].plot.bar(ax=ax, color=['C0','C1'])
        if row['corpus'] == 'fairview':
            corpus = 'Fairview'
            label =  '(d) '
        elif row['corpus'] == 'mipacq':
            corpus = 'MiPACQ'
            if ncols == 3:
                label = '(f) '
            else:
                label = '(b) '
        else:
            corpus = 'i2b2'
            label = '(e) '
        
        ax.set_title(label + corpus)
        ax.tick_params(axis='x', labelrotation = 45)
    
    if sg == 'all':
        sg = 'All groups'

    # Defining custom 'xlim' and 'ylim' values.
    custom_ylim = (0, 60)

    # Setting the values for all axes.
    plt.setp(axs, ylim=custom_ylim)
    fig.suptitle('Figure ' + str(ix) + ' ' + sg)

在上面的代碼中，我遍歷我的 df 抓取以下行以生成兩個單獨的子圖：

# BUT, I WANT TO PLOT BOTH ROWS ON SAME SUBPLOT
row[['mono p', 'non p']].plot.bar(ax=ax, color=['C0','C1'])
row[['plus p', 'minus p']].plot.bar(ax=ax, color=['C0','C1'])

無論我怎么做，我都無法在單個子圖中獲得所需的兩行（我總是得到一個空行，第二行沒有數據）。

Answer 1

• 查看內嵌評論
• 在python 3.8.12 、 pandas 1.3.3 、 matplotlib 3.4.3 、 seaborn 0.11.2

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns  # seaborn is a high-level api for matplotlib

# sample dataframe
data = {'corpus': ['fairview', 'i2b2', 'mipacq'], 'group': ['all', 'all', 'all'], 'mono p': [49, 46, 44], 'non p': [51, 54, 56], 'plus p': [49, 46, 43], 'minus p': [0, 0, 1]}
df = pd.DataFrame(data)

semgroups = df.group.unique()  # unique groups
corpus = df.corpus.unique()  # unique corpus
rows = [['mono p', 'non p'], ['plus p', 'minus p']]  # columns for each row of plots
for sg in semgroups:
    
    i = semgroups.index(sg)
    ix = i + 7

    ncols = len(corpus)  # 3 columns for the example
    nrows = len(rows)  # 2 rows for the example

    # create a figure with 2 rows of 3 columns: axes is a 2x3 array of <AxesSubplot:>
    fig, axes = plt.subplots(nrows, ncols, sharey=True, figsize=(12, 10))

    # iterate through each plot row combined with a list from rows
    for axe, row in zip(axes, rows):
        # iterate through each plot column of the current row
        for i, ax in enumerate(axe):
            
            # select the data for each plot
            data = df.loc[df.group.eq(sg) & df.corpus.eq(corpus[i]), row]
            
            # plot the dataframe, but setting the bar color is more difficult
#             data.T.plot(kind='bar', legend=False, ax=ax)
            
            # plot the data with seaborn, which is easier to color the bars
            sns.barplot(data=data, ax=ax)

            if corpus[i] == 'fairview':
                l2 = 'Fairview'
                l1 =  '(d) '
            elif corpus[i] == 'mipacq':
                l2 = 'MiPACQ'
                if ncols == 3:
                    l1 = '(f) '
                else:
                    l1 = '(b) '
            else:
                l2 = 'i2b2'
                l1 = '(e) '

            ax.set_title(l1 + l2)
            ax.tick_params(axis='x', labelrotation = 45)
    
    if sg == 'all':
        sg = 'All groups'

    # Defining custom 'xlim' and 'ylim' values.
    custom_ylim = (0, 60)

    # Setting the values for all axes.
    plt.setp(axes, ylim=custom_ylim)
    fig.suptitle('Figure ' + str(ix) + ' ' + sg)
    fig.tight_layout()
    plt.show()