CODEWITHSUNDEEP
pythonmatplotlibgraphseabornaverage
kg5425
kg5425

Reputation: 469

Seaborn/Matplotlib create custom error band along line

I have an experiment with 10 participants and 96 accuracies collected for each. acc_i represents the participant's overall accuracy at timestep i. Therefore, I have a 10x96 numpy matrix which looks like this:

[[acc_0,acc_1,acc_2,...acc_95]
 [acc_0,acc_1,acc_2,...acc_95]
 [acc_0,acc_1,acc_2,...acc_95]
            .
            .
            .
 [acc_0,acc_1,acc_2,...acc_95]]

I want to plot a line of the average accuracy among all participants at each timestep, along with an error band that shows the average +- 1 standard deviation. I can calculate the average and standard deviations separately using pd.Series(np.average(human_accuracies, axis=0)) and pd.Series(np.std(human_accuracies, axis=0)). However, this gives me two separate lines on a graph when I use:

sns.lineplot(data=avg_accuracies)
sns.lineplot(data=sd_accuracies)

This is shown below:

enter image description here

How can I make my plot into something more like what is shown here:

enter image description here

I'd like the error band to be calculated using the standard deviation at each time step i +- the average accuracy at each time step i.

Upvotes: 1

Views: 1210

Answers (1)

Quang Hoang
Quang Hoang

Reputation: 150785

You can use plt.fill_between like this:

plt.fill_between(x=np.arange(len(avg_accuracies)),
                 y1=avg_accuracies - sd_accuracies,
                 y2=avg_accuracies + sd_accuracies,
                 alpha=0.25
                 )
plt.plot(np.arange(len(avg_accuracies)), avg_accuracies)

Output:

enter image description here

Upvotes: 3

Related Questions