绘制直方图的线图

Question

我正在尝试尽可能多地使用 Altair 重现此图表。 https://fivethirtyeight.com/wp-content/uploads/2014/04/hickey-bechdel-11.png?w=575

我坚持让黑线划分通过/失败。这类似于这个 Altair 示例：https : //altair-viz.github.io/gallery/step_chart.html。但是：在 538 viz 中，最终日期的值必须扩展到最后一个元素的整个宽度。在步骤图示例和我的解决方案中，只要遇到最后一个日期元素，该行就会停止。

我查看了 altair 的 github 和 google 组，没有发现与此问题类似的内容。

import altair as alt
import pandas as pd

movies=pd.read_csv('https://raw.githubusercontent.com/fivethirtyeight/data/master/bechdel/movies.csv')
domain = ['ok', 'dubious','men', 'notalk', 'nowomen']

base=alt.Chart(movies).encode(
  alt.X("year:N",bin=alt.BinParams(step=5,extent=[1970,2015]),axis=alt.Axis(labelAngle=0, labelLimit=50,labelFontSize=8),title=None),  alt.Y("count()",stack='normalize',title=None,axis=alt.Axis(format='%',values=[0, 0.25,0.50,0.75,1]))

).properties(width=400)
main=base.transform_calculate(cleanrank='datum.clean_test == "ok" ? 1 : datum.clean_test == "dubious" ? 2 : datum.clean_test == "men" ? 3 : datum.clean_test == "notalk" ? 4 : 5'
                ).mark_bar(stroke='white' #add horizontal lines
                ).encode(  
  alt.Color("clean_test:N",scale=alt.Scale(
      domain=domain,
      range=['dodgerblue', 'skyblue', 'pink', 'coral','red']))
    ,order=alt.Order('cleanrank:O', sort='ascending')
)

extra=base.transform_calculate(cleanpass='datum.clean_test == "ok" ? "PASS" : datum.clean_test == "dubious" ? "PASS" : "FAIL"'
                      ).mark_line(interpolate='step-after'
                      ).encode(alt.Color("cleanpass:N",scale=alt.Scale(domain=['PASS','FAIL'],range=['black','white']))
                      )



alt.layer(main,extra).configure_scale(
    bandPaddingInner=0.01 #smaller vertical lines
).resolve_scale(color='independent')

Answer 1

使步骤图覆盖第一个 bin 的开头直到最后一个 bin 的末尾的一种相当老套的方法是手动控制 bin 位置（使用有序 bin 的排名）。

这样我们就可以添加两行：一行移动了一个 bin，'step-after'另一行step-before移动了一个 bin。从这里开始，刻度标签仍然需要替换并以适当的箱标签居中，例如来自pd.cut...的级别

数据框准备

import altair as alt
import pandas as pd

movies=pd.read_csv('https://raw.githubusercontent.com/fivethirtyeight/data/master/bechdel/movies.csv')
domain = ['ok', 'dubious','men', 'notalk', 'nowomen']

movies['year_bin'] = pd.cut(movies['year'], range(1970, 2016, 5))
movies['year_rank'] = movies['year_bin'].cat.codes
movies = movies[movies['year_rank']>=0]
df_plot = movies[['year_rank', 'clean_test']].copy()
df_plot['year_rank_end'] = df_plot['year_rank'] + 1
df_plot['clean_pass'] = df_plot['clean_test'].apply(lambda x: 'PASS' if x in ['ok', 'dubious'] else 'FAIL')

图表声明

base=alt.Chart(df_plot).encode(
    x=alt.X('year_rank', 
        axis=alt.Axis(labelAngle=0, labelLimit=50,labelFontSize=8),
        title=None
        ),  
  x2='year_rank_end',
  y=alt.Y('count()',title=None, stack='normalize',
        axis=alt.Axis(format='%',values=[0, 0.25,0.50,0.75,1])
        )
).properties(width=400)

main=base.transform_calculate(
    cleanrank='datum.clean_test == "ok" ? 1 : datum.clean_test == "dubious" ? 2 : datum.clean_test == "men" ? 3 : datum.clean_test == "notalk" ? 4 : 5'
    ).mark_bar(
        stroke='white' #add horizontal lines
    ).encode( 
  alt.Color("clean_test:N",scale=alt.Scale(
      domain=domain,
      range=['dodgerblue', 'skyblue', 'pink', 'coral','red']))
    ,order=alt.Order('cleanrank:O', sort='ascending')
)

extra=base.transform_calculate(
    ).mark_line(
        interpolate='step-after'
    ).encode(
        alt.Color("clean_pass:N",scale=alt.Scale(domain=['PASS','FAIL'],range=['black','white']))
    )

extra2=base.transform_calculate(
    # shift data by one bin, so that step-before matches the unshifted step-after
    year_rank='datum.year_rank +1' 
    ).mark_line(
        interpolate='step-before'
    ).encode(
        alt.Color("clean_pass:N",scale=alt.Scale(domain=['PASS','FAIL'],range=['black','white']), legend=None)
    )

alt.layer(main, extra, extra2).configure_scale(
    bandPaddingInner=0.01 #smaller vertical lines
).resolve_scale(color='independent')