Altair添加回归直线后出现未定义颜色图例且无法修改回归线颜色问题

Let's break down exactly what's going on here and how to fix both problems:

1. Why you're seeing that extra undefined color legend #

Your original chart test has a Color('GNI cat:O') encoding, which groups your scatter points by income category. When you use test.transform_regression(), Altair automatically inherits that grouping—it runs a separate regression for each GNI cat group, creating multiple regression lines. These lines get tied to your existing color legend, which can look like an "undefined" legend entry (or just clutter up your existing legend with line markers for each category).

2. Why mark_line(color="red") isn't working #

Altair prioritizes encoding-level color settings over mark-level ones. Since your regression line is inheriting the Color('GNI cat:O') encoding from test, that grouping color overrides any color you set directly in mark_line().

The Fixes #

We have two straightforward ways to resolve this, depending on whether you want a single global regression line or separate lines per category (it sounds like you want a single global line):

Option 1: Create a standalone regression chart (cleanest approach)

Instead of building the regression from your grouped test chart, create it directly from your raw sortedGNI data. This way it won't inherit any color grouping or existing encodings:

test = alt.Chart(sortedGNI).mark_circle().encode(
    alt.X('CO2 Emission:Q', title="CO2 Emission per Capita 2019"),
    alt.Y('Human Development Index:Q'),
    alt.Color('GNI cat:O', scale=alt.Scale(scheme='redyellowgreen'), sort=["Very High GNI", "High GNI", "Mid GNI", "Low GNI"], title=["Gross National Income ", "Per Capita"]),
    tooltip = ['Country:N', 'Gross National Income Per Capita:Q', 'Human Development Index:Q', 'CO2 Emission:Q']
)

# Build regression from raw data, no inherited encodings
regression_line = alt.Chart(sortedGNI).transform_regression(
    "CO2 Emission", 
    "Human Development Index", 
    method="pow"
).mark_line(color="red")

# Combine the charts
test + regression_line

This will give you one red global regression line, with only your original GNI cat legend showing up.

Option 2: Override grouping and color encoding on the existing test chart

If you prefer to build off the test chart directly, you can disable grouping and explicitly set the line color via the encoding layer (which takes priority):

test = alt.Chart(sortedGNI).mark_circle().encode(
    alt.X('CO2 Emission:Q', title="CO2 Emission per Capita 2019"),
    alt.Y('Human Development Index:Q'),
    alt.Color('GNI cat:O', scale=alt.Scale(scheme='redyellowgreen'), sort=["Very High GNI", "High GNI", "Mid GNI", "Low GNI"], title=["Gross National Income ", "Per Capita"]),
    tooltip = ['Country:N', 'Gross National Income Per Capita:Q', 'Human Development Index:Q', 'CO2 Emission:Q']
)

# Disable grouping with groupby=[] and set color via encoding
regression_line = test.transform_regression(
    "CO2 Emission", 
    "Human Development Index", 
    method="pow",
    groupby=[]  # Forces a single global regression line
).encode(
    color=alt.value("red")  # Overrides inherited color encoding
).mark_line()

test + regression_line

Either approach will get you that red regression line and eliminate the extra legend clutter.

内容的提问来源于stack exchange，提问作者Welly Wong