如何绘制长格式数据集中分类列的比率?

How does one plot the ratio of of a categorical column in a long form data set?

我有一个包含列 typedate 的长格式数据集。 type 有两个类别 goldsilver。我想按日期绘制两者的比率。为此,必须进行一系列转变。他们在 pandas

中看起来像这样
mock_df = df.groupby(["date"])["type"].value_counts().unstack()
mock_df["gs_ratio"] = mock_df["gold"]/mock_df["silver"]
mock_df

数据

import pandas

df = pd.DataFrame.from_records([
    {"date": "2020-04-20", "type": "gold"},
    {"date": "2020-04-20", "type": "silver"},
    {"date": "2020-04-20", "type": "silver"},
    {"date": "2020-04-21", "type": "gold"},
    {"date": "2020-04-21", "type": "gold"},
    {"date": "2020-04-21", "type": "silver"},
    {"date": "2020-04-22", "type": "gold"},
    {"date": "2020-04-22", "type": "silver"},
    {"date": "2020-04-22", "type": "silver"},
    {"date": "2020-04-22", "type": "silver"}
])

df

尝试过的代码:

alt.Chart(df).transform_joinaggregate(
    gs_count='count(type)',
    groupby=["date:T"]
).transform_pivot(
    'type',
    groupby=['date:T'],
    value='gs_count'
).transform_calculate(
    gs_ratio="datum.gold/datum.silver"
).mark_line().encode(
    x='date:T',
    y="gs_ratio:Q"
)

您的方法存在一些问题:

  • 您不能在转换中使用类型简写。所以你应该使用列的实际名称,"date" 而不是 "date:T"
  • count(type) 不等同于 df.type.value_counts()。你应该做的是使用 count()type.
  • 分组
  • 使用transform_aggregate而不是transform_joinaggregate

放在一起:

alt.Chart(df).transform_aggregate(
    gs_count='count()',
    groupby=["date", "type"]
).transform_pivot(
    'type',
    groupby=['date'],
    value='gs_count'
).transform_calculate(
    gs_ratio="datum.gold/datum.silver"
).mark_line().encode(
    x='date:T',
    y="gs_ratio:Q"
)