python - 在 python 中使用 seaborn.PairGrid 生成相关矩阵时对角线上直方图的标题

Question

在花了一些时间了解 PairGrid 功能的工作原理之后，我已经快成功了。下面是生成我想要的绘图的代码，但 histfunc 中缺少一个小细节。我想要的是对角线上绘制的直方图的标题。如何将数据框列名称传递给 histfunc？任何想法表示赞赏。

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
import dcor
import random
from scipy.stats import linregress
from matplotlib import rc

font = {'family' : 'normal',
        'weight' : 'normal',
        'size'   : 4}
rc('font', **font)   

def corrmat(data):
    def cm2inch(value):
        """helper function for plotting. Converts cm to inch"""
        return value/2.54

    def dist_corr(X, Y, pval=True, nruns=100):
        """ Distance correlation with p-value from bootstrapping"""
        dc = dcor.distance_correlation(X, Y)
        pv = dcor.independence.distance_covariance_test(X, Y, exponent=1.0, num_resamples=nruns)[0]
        if pval:
            return (dc, pv)
        else:
            return dc    

    def linreg(X, Y, pval=True):
        """ Linear regression"""
        r2 = linregress(X,Y)[2]**2
        pv = linregress(X,Y)[3]
        if pval:
            return (r2, pv)
        else:
            return r2               

    def scatterfunc(x, y, **kws):
        # scatterplot with spline of deg=5 in red
        plt.scatter(x, y, linewidths=1, facecolor="k", s=10, alpha = 0.5)
        spline = np.polyfit(x, y, 5)
        model = np.poly1d(spline)
        x = np.sort(x)
        plt.plot(x,model(x),'r-')

    def histfunc(x, **kws):
        #  histogram
        plt.hist(x,bins=30,color = "black", ec="white")    
        """
        vvvvvvvvvvvvvvvvvvvv
        here something like 
        plt.title(label) 
        is missing but the **kws only contain label as string not as 
        parameter contaning the column name
        ^^^^^^^^^^^^^^^^^^^
        """

    def corrfunc(x, y, dc=False, **kws):  
        # different sizes, text anc color in relation to r/d values         
        if dc:
            d, p = dist_corr(x,y) 
        else:    
            d, p = linreg(x,y)

        if d<0.25:
            pclr = 'Black'
            fontsize = 16
        elif (d>=0.25) and (d<0.5):
            pclr = 'Blue'
            fontsize = 20
        elif (d>=0.5) and (p<0.75):
            pclr = 'Orange'
            fontsize = 25
        elif (p>0.75):
            pclr = 'Red'
            fontsize = 30

        if p<0.001:
            ptext = "***"
        elif (p>=0.001) and (p<0.01):
            ptext = "**"
        elif (p>=0.01) and (p<0.05):
            ptext = "*"
        elif (p>0.05):
            ptext = "n.sig"

        ax = plt.gca()
        if dc:
            ax.annotate(''.join(['DC: ',str(np.round(d,2)),'\n\n    ',ptext]),
                        xy=(0.3, 0.3), 
                        xycoords=ax.transAxes, 
                        color = pclr, 
                        fontsize = fontsize)
        else:
            ax.annotate(''.join(['r2: ',str(np.round(d,2)),'\n\n    ',ptext]),
                        xy=(0.3, 0.3), 
                        xycoords=ax.transAxes, 
                        color = pclr, 
                        fontsize = fontsize)

        plt.axis('off')


    plt.figure(num=None, figsize=(cm2inch(15), cm2inch(10)), dpi=300, facecolor='w', edgecolor='k')
    g = sns.PairGrid(data, diag_sharey=False)
    g.map_upper(scatterfunc)
    g.map_diag(histfunc)
    g.map_lower(corrfunc)
    plt.tight_layout()
    plt.show()


########

data = pd.DataFrame(np.random.random([1000,10]),columns=[str(i) for i in range(10)])   
for (i,col) in enumerate(data):
    if i > 1:
        if np.random.random()>0.5:
            data[col]= data[col] * data.iloc[:,random.sample(set(np.arange(0,i)),1 )[0]]
corrmat(data)

它生成的是

Answer 1

感谢 @ImportanceOfBeingErnest 在这里评论了更新的脚本，供那些可能觉得有用的人使用。我还将散点图切换为“较低”，以便轴标签可见。

import matplotlib.pyplot as plt
import seaborn as sns
import pandas as pd
import numpy as np
import dcor
import random
from scipy.stats import linregress
from matplotlib import rc

font = {'family' : 'normal',
        'weight' : 'normal',
        'size'   : 16}
rc('font', **font)   

def corrmat(data):
    def cm2inch(value):
        """helper function for plotting. Converts cm to inch"""
        return value/2.54

    def dist_corr(X, Y, pval=True, nruns=100):
        """ Distance correlation with p-value from bootstrapping"""
        dc = dcor.distance_correlation(X, Y)
        pv = dcor.independence.distance_covariance_test(X, Y, exponent=1.0, num_resamples=nruns)[0]
        if pval:
            return (dc, pv)
        else:
            return dc    

    def linreg(X, Y, pval=True):
        """ Linear regression"""
        r2 = linregress(X,Y)[2]**2
        pv = linregress(X,Y)[3]
        if pval:
            return (r2, pv)
        else:
            return r2               

    def scatterfunc(x, y, **kws):
        """ scatterplot with spline of deg=5 in red"""
        plt.scatter(x, y, linewidths=1, facecolor="k", s=10, alpha = 0.5)
        spline = np.polyfit(x, y, 5)
        model = np.poly1d(spline)
        x = np.sort(x)
        plt.plot(x,model(x),'r-')

    def histfunc(x, **kws):
        """ histogram"""
        plt.hist(x,bins=30,color = "black", ec="white")    

    def corrfunc(x, y, dc=False, **kws):  
        """different sizes, text anc color in relation to r/d values
           the dc parameter determines wheter distance correlation or 
           linear regression should be applied"""
        if dc:
            d, p = dist_corr(x,y) 
        else:    
            d, p = linreg(x,y)

        if d<0.25:
            pclr = 'Black'
            fontsize = 16
        elif (d>=0.25) and (d<0.5):
            pclr = 'Blue'
            fontsize = 20
        elif (d>=0.5) and (p<0.75):
            pclr = 'Orange'
            fontsize = 25
        elif (p>0.75):
            pclr = 'Red'
            fontsize = 30

        if p<0.001:
            ptext = "***"
        elif (p>=0.001) and (p<0.01):
            ptext = "**"
        elif (p>=0.01) and (p<0.05):
            ptext = "*"
        elif (p>0.05):
            ptext = "n.sig"

        ax = plt.gca()
        if dc:
            ax.annotate(''.join(['DC: ',str(np.round(d,2)),'\n\n    ',ptext]),
                        xy=(0.3, 0.3), 
                        xycoords=ax.transAxes, 
                        color = pclr, 
                        fontsize = fontsize)
        else:
            ax.annotate(''.join(['r2: ',str(np.round(d,2)),'\n\n    ',ptext]),
                        xy=(0.3, 0.3), 
                        xycoords=ax.transAxes, 
                        color = pclr, 
                        fontsize = fontsize)

        plt.axis('off')

    def make_diag_titles(g,titles):
        for (i,row) in enumerate(g.axes):
            g.axes[i][i].title.set_text(titles[i])
        return g
    ###
    # here the plot is put together
    plt.figure(num=None, figsize=(cm2inch(15), cm2inch(10)), dpi=300, facecolor='w', edgecolor='k')
    g = sns.PairGrid(data, diag_sharey=False)
    g.map_lower(scatterfunc)
    g.map_diag(histfunc)
    g.map_upper(corrfunc)
    g = make_diag_titles(g, data.columns)
    plt.tight_layout()
    plt.show()


########

data = pd.DataFrame(np.random.random([1000,10]),columns=[str(i) for i in range(10)])   
for (i,col) in enumerate(data):
    if i > 1:
        if np.random.random()>0.5:
            data[col]= data[col] * data.iloc[:,random.sample(set(np.arange(0,i)),1 )[0]]
corrmat(data)