97
votes

I'm using Seaborn's lmplot to plot a linear regression, dividing my dataset into two groups with a categorical variable.

For both x and y, I'd like to manually set the lower bound on both plots, but leave the upper bound at the Seaborn default. Here's a simple example:

import pandas as pd
import seaborn as sns
import random

n = 200
random.seed(2014)
base_x = [random.random() for i in range(n)]
base_y = [2*i for i in base_x]
errors = [random.uniform(0,1) for i in range(n)]
y = [i+j for i,j in zip(base_y,errors)]

df = pd.DataFrame({'X': base_x,
                   'Y': y,
                   'Z': ['A','B']*(n/2)})

mask_for_b = df.Z == 'B'
df.loc[mask_for_b,['X','Y']] = df.loc[mask_for_b,] *2

sns.lmplot('X','Y',df,col='Z',sharex=False,sharey=False)

This outputs the following: enter image description here

But in this example, I'd like the xlim and the ylim to be (0,*) . I tried using sns.plt.ylim and sns.plt.xlim but those only affect the right-hand plot. Example:

sns.plt.ylim(0,)
sns.plt.xlim(0,)

enter image description here

How can I access the xlim and ylim for each plot in the FacetGrid?

2
By the way, if you familiarize yourself with the numpy.random module, you can save yourself a lot of time generating random data (which can be a very useful thing to do!). For example, you could get base_x and base_y with base_x = np.random.rand(n); base_y = base_x * 2. The y variable can then be similarly generated with vectorized operations. - mwaskom

2 Answers

162
votes

The lmplot function returns a FacetGrid instance. This object has a method called set, to which you can pass key=value pairs and they will be set on each Axes object in the grid.

Secondly, you can set only one side of an Axes limit in matplotlib by passing None for the value you want to remain as the default.

Putting these together, we have:

g = sns.lmplot('X', 'Y', df, col='Z', sharex=False, sharey=False)
g.set(ylim=(0, None))

enter image description here

85
votes

You need to get hold of the axes themselves. Probably the cleanest way is to change your last row:

lm = sns.lmplot('X','Y',df,col='Z',sharex=False,sharey=False)

Then you can get hold of the axes objects (an array of axes):

axes = lm.axes

After that you can tweak the axes properties

axes[0,0].set_ylim(0,)
axes[0,1].set_ylim(0,)

creates:

enter image description here