地块分布

Question

I'm having a pandas dataset that contains an integer and a float value:

>>> df2[['AGE_REF', 'RETSURV']].dtypes
AGE_REF      int64
RETSURV    float64
dtype: object

I'd like to plot the joint distribution using pandas. I didn't see a simple way of pandas visualizing the joint distribution, but I stumbled across seaborn . So I tried to adjust code that I already found for my purposes:

>>> import seaborn as sns
>>> sns.jointplot('AGE_REF', "RETSURV", df2,
              kind="hex")
Traceback (most recent call last):
  File "<input>", line 2, in <module>
  File "/usr/local/lib/python2.7/site-packages/seaborn/distributions.py", line 969, in jointplot
    gridsize = int(np.mean([x_bins, y_bins]))
OverflowError: cannot convert float infinity to integer

I found a related bug report , so I tried to follow the workaround there - without success:

>>> sns.jointplot('AGE_REF', "RETSURV", df2,
              kind="hex", marginal_kws={"bins": 10})
Traceback (most recent call last):
  File "<input>", line 2, in <module>
  File "/usr/local/lib/python2.7/site-packages/seaborn/distributions.py", line 969, in jointplot
    gridsize = int(np.mean([x_bins, y_bins]))
OverflowError: cannot convert float infinity to integer

Answer 1

The default hexbin gridsize uses the same reference rule calculation as the histograms, so you'll need to set that directly too if you have data that violate those assumptions somehow:

sns.jointplot(x, y, kind="hex",
              joint_kws={"gridsize": 10},
              marginal_kws={"bins": 10})