Seaborn 点图并不意味着

Question

我喜欢这个 seaborn example，并希望使用 FacetGrid 将其应用到 pandas 数据框以比较不同的场景：

df_new = pd.read_json('{"TA":{"229":-30.0,"230":-30.0,"192":23.0,"193":23.0,"248":60.0,"249":60.0,"126":-30.0,"127":-30.0,"88":23.0,"89":23.0,"150":60.0,"151":60.0,"239":-30.0,"240":-30.0,"197":23.0,"198":23.0,"256":60.0,"257":60.0,"135":-30.0,"136":-30.0,"94":23.0,"95":23.0,"164":60.0,"165":60.0,"438":-30.0,"439":-30.0,"291":23.0,"405":23.0,"453":60.0,"454":60.0,"341":-30.0,"342":-30.0,"292":23.0,"293":23.0,"365":60.0,"366":60.0,"445":-30.0,"446":-30.0,"410":23.0,"411":23.0,"462":60.0,"463":60.0,"357":-30.0,"358":-30.0,"297":23.0,"298":23.0,"371":60.0,"372":60.0},"Type":{"229":"A","230":"A","192":"A","193":"A","248":"A","249":"A","126":"P","127":"P","88":"P","89":"P","150":"P","151":"P","239":"A","240":"A","197":"A","198":"A","256":"A","257":"A","135":"P","136":"P","94":"P","95":"P","164":"P","165":"P","438":"A","439":"A","291":"A","405":"A","453":"A","454":"A","341":"P","342":"P","292":"P","293":"P","365":"P","366":"P","445":"A","446":"A","410":"A","411":"A","462":"A","463":"A","357":"P","358":"P","297":"P","298":"P","371":"P","372":"P"},"Value":{"229":57.36232,"230":52.97104,"192":59.82472,"193":56.70568,"248":72.30088,"249":68.56624,"126":71.68528,"127":79.15456,"88":84.1204,"89":82.2736,"150":77.26672,"151":81.00136,"239":70.41304,"240":82.2736,"197":76.03552,"198":83.5048,"256":82.8892,"257":88.51168,"135":89.74288,"136":97.21216,"94":99.1,"95":95.98096,"164":95.98096,"165":96.59656,"438":64.8316,"439":73.53208,"291":107.18488,"405":82.2736,"453":77.26672,"454":86.00824,"341":105.29704,"342":97.21216,"292":108.41608,"293":100.3312,"365":84.77704,"366":88.51168,"445":46.11736,"446":52.35544,"410":62.32816,"411":65.4472,"462":71.06968,"463":74.80432,"357":77.92336,"358":79.15456,"297":94.09312,"298":87.23944,"371":82.2736,"372":98.4844},"Group":{"229":"FA","230":"FA","192":"FA","193":"FA","248":"FA","249":"FA","126":"FA","127":"FA","88":"FA","89":"FA","150":"FA","151":"FA","239":"FB","240":"FB","197":"FB","198":"FB","256":"FB","257":"FB","135":"FB","136":"FB","94":"FB","95":"FB","164":"FB","165":"FB","438":"RB","439":"RB","291":"RB","405":"RB","453":"RB","454":"RB","341":"RB","342":"RB","292":"RB","293":"RB","365":"RB","366":"RB","445":"RC","446":"RC","410":"RC","411":"RC","462":"RC","463":"RC","357":"RC","358":"RC","297":"RC","298":"RC","371":"RC","372":"RC"}}')
g = sns.factorplot(x="Value", y="Type", hue="TA",
                   col="Group", data=df_new, col_wrap=2,
                   kind="strip", dodge=True, jitter=True, alpha=.5)
g = g.map_dataframe(sns.pointplot, x="Value", y="Type", hue="TA",
                    dodge=.532, join=False, palette="dark", markers="d", scale=.75, ci=None)

def myplot(x, y, **kwargs):
    ax = plt.gca()
    data = kwargs.pop("data")
    print(data.shape, "in plotting group", data.iloc[0]['Group'])
    groups = data.groupby([y, 'TA'])
    for label, group_df in groups:
        print("Group label:", label, "Group mean: {:.2f}".format(group_df[x].mean()))

g = g.map_dataframe(myplot, x="Value", y="Type")
g.set_titles(row_template="{row_name}", col_template="{col_name}")

问题是seaborn的点图提供的平均值对于绘制组RB是错误的。

出于调试目的，我添加了一个自定义函数myplot，它只输出每个绘图组的数据及其平均值：

((12, 4), 'in plotting group', u'FA')
('Group label:', (u'A', -30), 'Group mean: 55.17')
('Group label:', (u'A', 23), 'Group mean: 58.27')
('Group label:', (u'A', 60), 'Group mean: 70.43')
('Group label:', (u'P', -30), 'Group mean: 75.42')
('Group label:', (u'P', 23), 'Group mean: 83.20')
('Group label:', (u'P', 60), 'Group mean: 79.13')
((12, 4), 'in plotting group', u'FB')
('Group label:', (u'A', -30), 'Group mean: 76.34')
('Group label:', (u'A', 23), 'Group mean: 79.77')
('Group label:', (u'A', 60), 'Group mean: 85.70')
('Group label:', (u'P', -30), 'Group mean: 93.48')
('Group label:', (u'P', 23), 'Group mean: 97.54')
('Group label:', (u'P', 60), 'Group mean: 96.29')
((12, 4), 'in plotting group', u'RB')
('Group label:', (u'A', -30), 'Group mean: 69.18')
('Group label:', (u'A', 23), 'Group mean: 94.73')
('Group label:', (u'A', 60), 'Group mean: 81.64')
('Group label:', (u'P', -30), 'Group mean: 101.25')
('Group label:', (u'P', 23), 'Group mean: 104.37')
('Group label:', (u'P', 60), 'Group mean: 86.64')
((12, 4), 'in plotting group', u'RC')
('Group label:', (u'A', -30), 'Group mean: 49.24')
('Group label:', (u'A', 23), 'Group mean: 63.89')
('Group label:', (u'A', 60), 'Group mean: 72.94')
('Group label:', (u'P', -30), 'Group mean: 78.54')
('Group label:', (u'P', 23), 'Group mean: 90.67')
('Group label:', (u'P', 60), 'Group mean: 90.38')

所以我在这里看到的是计算的平均值与点图不对应。 我的计算错了吗？我是否为绘图功能设置了错误的参数？

Answer 1

可以看出"P" 和"A" 类型的方法在左下角的子图中互换了。

虽然factorplot 本身确保在其子图中具有相同的顺序，但映射的pointplot 不知道此顺序。

为了确保在任何地方都使用相同的顺序，您需要将此顺序提供给绘图函数。

g = sns.factorplot(..., order=["A","P"])
g.map_dataframe(sns.pointplot, ..., order=["A","P"]))

为了安全起见，hue_order 也可以指定为hue_order=[-30,23,60]。

完整示例：

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

df_new = pd.read_json('{"TA":{"229":-30.0,"230":-30.0,"192":23.0,"193":23.0,"248":60.0,"249":60.0,"126":-30.0,"127":-30.0,"88":23.0,"89":23.0,"150":60.0,"151":60.0,"239":-30.0,"240":-30.0,"197":23.0,"198":23.0,"256":60.0,"257":60.0,"135":-30.0,"136":-30.0,"94":23.0,"95":23.0,"164":60.0,"165":60.0,"438":-30.0,"439":-30.0,"291":23.0,"405":23.0,"453":60.0,"454":60.0,"341":-30.0,"342":-30.0,"292":23.0,"293":23.0,"365":60.0,"366":60.0,"445":-30.0,"446":-30.0,"410":23.0,"411":23.0,"462":60.0,"463":60.0,"357":-30.0,"358":-30.0,"297":23.0,"298":23.0,"371":60.0,"372":60.0},"Type":{"229":"A","230":"A","192":"A","193":"A","248":"A","249":"A","126":"P","127":"P","88":"P","89":"P","150":"P","151":"P","239":"A","240":"A","197":"A","198":"A","256":"A","257":"A","135":"P","136":"P","94":"P","95":"P","164":"P","165":"P","438":"A","439":"A","291":"A","405":"A","453":"A","454":"A","341":"P","342":"P","292":"P","293":"P","365":"P","366":"P","445":"A","446":"A","410":"A","411":"A","462":"A","463":"A","357":"P","358":"P","297":"P","298":"P","371":"P","372":"P"},"Value":{"229":57.36232,"230":52.97104,"192":59.82472,"193":56.70568,"248":72.30088,"249":68.56624,"126":71.68528,"127":79.15456,"88":84.1204,"89":82.2736,"150":77.26672,"151":81.00136,"239":70.41304,"240":82.2736,"197":76.03552,"198":83.5048,"256":82.8892,"257":88.51168,"135":89.74288,"136":97.21216,"94":99.1,"95":95.98096,"164":95.98096,"165":96.59656,"438":64.8316,"439":73.53208,"291":107.18488,"405":82.2736,"453":77.26672,"454":86.00824,"341":105.29704,"342":97.21216,"292":108.41608,"293":100.3312,"365":84.77704,"366":88.51168,"445":46.11736,"446":52.35544,"410":62.32816,"411":65.4472,"462":71.06968,"463":74.80432,"357":77.92336,"358":79.15456,"297":94.09312,"298":87.23944,"371":82.2736,"372":98.4844},"Group":{"229":"FA","230":"FA","192":"FA","193":"FA","248":"FA","249":"FA","126":"FA","127":"FA","88":"FA","89":"FA","150":"FA","151":"FA","239":"FB","240":"FB","197":"FB","198":"FB","256":"FB","257":"FB","135":"FB","136":"FB","94":"FB","95":"FB","164":"FB","165":"FB","438":"RB","439":"RB","291":"RB","405":"RB","453":"RB","454":"RB","341":"RB","342":"RB","292":"RB","293":"RB","365":"RB","366":"RB","445":"RC","446":"RC","410":"RC","411":"RC","462":"RC","463":"RC","357":"RC","358":"RC","297":"RC","298":"RC","371":"RC","372":"RC"}}')

g = sns.factorplot(x="Value", y="Type", hue="TA",
                   col="Group", data=df_new, col_wrap=2,
                   kind="strip", dodge=True, palette="dark",jitter=True, alpha=.5, 
                   order=["A","P"], hue_order=[-30,23,60])
g = g.map_dataframe(sns.pointplot, x="Value", y="Type", hue="TA", 
                    order=["A","P"], hue_order=[-30,23,60],
                    dodge=.532, join=False, palette="dark", markers="d", scale=.75, ci=None)

g.set_titles(row_template="{row_name}", col_template="{col_name}")

plt.show()