直方圖的畫線 plot

Question

我正在嘗試盡可能多地使用 Altair 重現此圖表。 https://fivethirtyeight.com/wp-content/uploads/2014/04/hickey-bechdel-11.png?w=575

我堅持讓黑線划分通過/失敗。 這類似於 Altair 示例： https://altair-viz.github.io/gallery/step_chart.html 。 但是：在 538 可視化項中，最終日期的值必須擴展到最后一個元素的整個寬度。 在步驟圖示例和我的解決方案中，該行在遇到最后一個日期元素后立即停止。

我查看了 altair 的 github 和 google 組，沒有發現與此問題類似的問題。

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

一種 - 相當 hacky - 使步驟圖覆蓋第一個箱子的開頭直到最后一個箱子結束的方法是手動控制箱子位置（使用有序箱子的等級）。

通過這種方式，我們可以添加兩行：一行帶有'step-after' ，另一行帶有step-before移動一個 bin。 從這里開始，刻度標簽仍然需要替換並以適當的 bin 標簽居中，例如來自pd.cut的級別 ...

Dataframe 准備

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')

圖表聲明

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')