What is the difference between these seaborn.boxplot implementations?

Question

My current prof entered this code in our tutorial jupyter notebook:

f = plt.figure(figsize=(24, 4))  
sb.boxplot(data = hp, orient = "h")

but in previous year’s tutorial videos, this was the code:

f, axes = plt.subplots(1, 1, figsize=(24,4))  
sb.boxplot(hp, orient = "h")

I have tried to run the 2nd code but it shows up as an error

The truth value of a Series is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().

This was the error that I got. I am not sure why it can run on the previous year’s tutorial videos but can’t on my notebook.

Asked By: Clarissa Chen

||

Source

Answer 1

Your error is not about matplotlib, but seaborn. You probably use a 0.11 version.

The boxplot function signature for 0.11 series is:

sns.boxplot(*, x=None, y=None, hue=None, data=None, ..., **kwargs)

So when you use sns.boxplot(hp, orient='h'), hp is the value for x and not data parameter. The first version of your code works because you explicitly named the data parameter.

Since 0.12 version, the signature is:

sns.boxplot(data=None, *, x=None, y=None, ..., **kwargs)

The only unnamed argument allowed is data, so both of your code will work.

Solution: update to Seaborn 0.12 or use data=hp as parameter of boxplot (and other functions).

Summary:

# version 0.11
sns.boxplot(data=hp, orient='h')  # works
sns.boxplot(hp, orient='h')       # failed

# version 0.12
sns.boxplot(data=hp, orient='h')  # works
sns.boxplot(hp, orient='h')       # works

Answered By: Corralien

What is the difference between these seaborn.boxplot implementations?

Question:

Answers: