Question

是否可以在Python中创建Bland-Altman plot？我似乎无法找到任何相关信息。

此类情节的另一个名称是 Tukey均值差异。

示例：

enter image description here

Answer 1

如果我正确理解了情节背后的理论，这段代码应该提供基本的绘图，而你可以根据自己的特殊需要进行配置。

import matplotlib.pyplot as plt
import numpy as np

def bland_altman_plot(data1, data2, *args, **kwargs):
    data1     = np.asarray(data1)
    data2     = np.asarray(data2)
    mean      = np.mean([data1, data2], axis=0)
    diff      = data1 - data2                   # Difference between data1 and data2
    md        = np.mean(diff)                   # Mean of the difference
    sd        = np.std(diff, axis=0)            # Standard deviation of the difference

    plt.scatter(mean, diff, *args, **kwargs)
    plt.axhline(md,           color='gray', linestyle='--')
    plt.axhline(md + 1.96*sd, color='gray', linestyle='--')
    plt.axhline(md - 1.96*sd, color='gray', linestyle='--')

data1和data2中的相应元素用于计算绘制点的坐标。

然后你可以通过运行例如

创建一个图

from numpy.random import random

bland_altman_plot(random(10), random(10))
plt.title('Bland-Altman Plot')
plt.show()

Bland-Altman Plot

Answer 2

现在已在statsmodels中实现：https://www.statsmodels.org/devel/generated/statsmodels.graphics.agreement.mean_diff_plot.html

这是他们的例子：

import statsmodels.api as sm
import numpy as np
import matplotlib.pyplot as plt

# Seed the random number generator.
# This ensures that the results below are reproducible.
np.random.seed(9999)
m1 = np.random.random(20)
m2 = np.random.random(20)

f, ax = plt.subplots(1, figsize = (8,5))
sm.graphics.mean_diff_plot(m1, m2, ax = ax)

plt.show()

产生以下结果：

Answer 3

我接受了sodd的回答，并做了一个有计划的实施。这似乎是轻松共享它的最佳位置。

from scipy.stats import linregress
import numpy as np
import plotly.graph_objects as go
def bland_altman_plot(data1, data2, data1_name='A', data2_name='B', subgroups=None, plotly_template='none', annotation_offset=0.05, plot_trendline=True, n_sd=1.96,*args, **kwargs):
    data1 = np.asarray( data1 )
    data2 = np.asarray( data2 )
    mean = np.mean( [data1, data2], axis=0 )
    diff = data1 - data2  # Difference between data1 and data2
    md = np.mean( diff )  # Mean of the difference
    sd = np.std( diff, axis=0 )  # Standard deviation of the difference


    fig = go.Figure()

    if plot_trendline:
        slope, intercept, r_value, p_value, std_err = linregress(mean, diff)
        trendline_x = np.linspace(mean.min(), mean.max(), 10)
        fig.add_trace(go.Scatter(x=trendline_x, y=slope*trendline_x + intercept,
                                 name='Trendline',
                                 mode='lines',
                                 line=dict(
                                        width=4,
                                        dash='dot')))
    if subgroups is None:
        fig.add_trace( go.Scatter( x=mean, y=diff, mode='markers', **kwargs))
    else:
        for group_name in np.unique(subgroups):
            group_mask = np.where(np.array(subgroups) == group_name)
            fig.add_trace( go.Scatter(x=mean[group_mask], y=diff[group_mask], mode='markers', name=str(group_name), **kwargs))



    fig.add_shape(
        # Line Horizontal
        type="line",
        xref="paper",
        x0=0,
        y0=md,
        x1=1,
        y1=md,
        line=dict(
            # color="Black",
            width=6,
            dash="dashdot",
        ),
        name=f'Mean {round( md, 2 )}',
    )
    fig.add_shape(
        # borderless Rectangle
        type="rect",
        xref="paper",
        x0=0,
        y0=md - n_sd * sd,
        x1=1,
        y1=md + n_sd * sd,
        line=dict(
            color="SeaGreen",
            width=2,
        ),
        fillcolor="LightSkyBlue",
        opacity=0.4,
        name=f'±{n_sd} Standard Deviations'
    )

    # Edit the layout
    fig.update_layout( title=f'Bland-Altman Plot for {data1_name} and {data2_name}',
                       xaxis_title=f'Average of {data1_name} and {data2_name}',
                       yaxis_title=f'{data1_name} Minus {data2_name}',
                       template=plotly_template,
                       annotations=[dict(
                                        x=1,
                                        y=md,
                                        xref="paper",
                                        yref="y",
                                        text=f"Mean {round(md,2)}",
                                        showarrow=True,
                                        arrowhead=7,
                                        ax=50,
                                        ay=0
                                    ),
                                   dict(
                                       x=1,
                                       y=n_sd*sd + md + annotation_offset,
                                       xref="paper",
                                       yref="y",
                                       text=f"+{n_sd} SD",
                                       showarrow=False,
                                       arrowhead=0,
                                       ax=0,
                                       ay=-20
                                   ),
                                   dict(
                                       x=1,
                                       y=md - n_sd *sd + annotation_offset,
                                       xref="paper",
                                       yref="y",
                                       text=f"-{n_sd} SD",
                                       showarrow=False,
                                       arrowhead=0,
                                       ax=0,
                                       ay=20
                                   ),
                                   dict(
                                       x=1,
                                       y=md + n_sd * sd - annotation_offset,
                                       xref="paper",
                                       yref="y",
                                       text=f"{round(md + n_sd*sd, 2)}",
                                       showarrow=False,
                                       arrowhead=0,
                                       ax=0,
                                       ay=20
                                   ),
                                   dict(
                                       x=1,
                                       y=md - n_sd * sd - annotation_offset,
                                       xref="paper",
                                       yref="y",
                                       text=f"{round(md - n_sd*sd, 2)}",
                                       showarrow=False,
                                       arrowhead=0,
                                       ax=0,
                                       ay=20
                                   )
                               ])
    return fig

Answer 4

也许我错过了什么，但这看起来很简单：

from numpy.random import random
import matplotlib.pyplot as plt

x = random(25)
y = random(25)

plt.title("FooBar")
plt.scatter(x,y)
plt.axhline(y=0.5,linestyle='--')
plt.show()

这里我只是在0和1之间创建一些随机数据，然后我在y = 0.5处随机放置一条水平线 - 但是你可以随心所欲地放置任意数量。

Answer 5

pyCompare 有 Bland-Altman 图（见 demo from Jupyter）

import pyCompare
method1 = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20]
method2 = [1.03, 2.05, 2.79, 3.67, 5.00, 5.82, 7.16, 7.69, 8.53, 10.38, 11.11, 12.17, 13.47, 13.83, 15.15, 16.12, 16.94, 18.09, 19.13, 19.54]
pyCompare.blandAltman(method1, method2)

pyCompare 模块的详细信息 in PyPI

最终产品看起来像：

Python中的Bland-Altman情节

5 个答案: