如何将一个分类列的色调添加到 Seaborn 条形图（每列的平均值和标准）？

Question

sns.barplot(data = df, ci='sd') gives a barplot of each column mean with std like below.

Let's say there is one column named 'category', which has 3 kinds of values A, B, and C.

I want to add a hue parameter to the barplot from this column in order to split each bar of column mean into three according to the category. Like this:

If I directly add hue = 'category' , I got an error Cannot use `hue` without `x` and `y` .

What should x and y be here?

Answer 1

According to the docs, it looks like if you don't pass x or y , then your DataFrame will be interpreted as 'wide-form', and each numeric column will be plotted. Then if you want to maintain the plot layout you have, but use hue='category' , you'll need to convert your DataFrame to 'long-form' using melt , using the former columns -- now the var_name column -- as x . Without the specifics, I can't say exactly which columns you'll need to use, but it should look roughly like this:

melted_df = df.melt(var_name='variable', value_name='value')
sns.barplot(melted_df, x='variable', y='value', hue='category')

Does this work for you?

Note that in general, Seaborn expects your data in long format. I think you'll find it makes most operations more convenient, as each "unit" of the plot is expressed as a group in a groupby, which is simpler than dealing with grouping along both the row and column axes.

Answer 2

If you are ok with separate subplots for each column, one way of doing this would be like:

fig, ax = plt.subplots(1, len(df.columns), figsize=(120,6))
for i in range(len(df.columns)):
    sns.barplot(x='category', y=df.columns[i], ax=ax[i], data=df, ci='sd')

Does this work for you?