Search code examples
pythonpandasdataframeseaborn

Problem plotting pandas dataframe containing arrays

I have a tricky question for you concerning data structure in pandas for plotting with seaborn.

Let's imagine, I have several experiments, each of them performed in different conditions. The result of each experiment is an array with a few thousand floats.

I was considering to have all the experiment results stored in a single pandas dataframe, in the so-called long-format, i.e. each row is one experiment, and each column is a variable. Almost all the variables are used to define the experimental conditions and then one variable containing the array of float with experiment results.

Something like this:

df = pd.DataFrame({'id':[1,2], 'temp':[21,22], 'oven':[0,1], 'values':[[1,2,3,4,5], [10,11,12,12,15,16,17]]})

So far so good.

Now I would like to use seaborn to make some plots. Imagine I want to plot an histogram of the values using id as a category.

I would do:

sns.histplot(df, x='values', hue='id')

But if I do so, I get an error message complaining that list is an unhashable type.

As a workaround, I changed the data structure, so that I have a row for each of the floats in the experiment results, but this is making the table unnecessarily huge.

Do you have any suggestion for me?

Solution

  • A flat/tidy DataFrame would be almost the same, but having one row per point.

    If you explode your dataset, this would work:

    df_flat = df.explode('values')

sns.histplot(df_flat, x='values', hue='id')

    Output:

    enter image description here

    Another option, would be to build a dictionary:

    sns.histplot(dict(zip(df['id'], df['values'])))

# or
sns.histplot({k: v.squeeze() for k,v in df.groupby('id')['values']})

    Output:

    enter image description here

    Finally, you can always plot manually:

    ax = plt.subplot()

for row in df.index:
    ax.hist(df.loc[row, 'values'], label=df.loc[row, 'id'])
ax.legend()

    enter image description here