Hist#
import hvplot.pandas # noqa
# hvplot.extension("matplotlib")
hist
is often a good way to start looking at continuous data to get a sense of the distribution. Similar methods include kde
(also available as density
).
from bokeh.sampledata.autompg import autompg_clean
autompg_clean.sample(n=5)
mpg | cyl | displ | hp | weight | accel | yr | origin | name | mfr | |
---|---|---|---|---|---|---|---|---|---|---|
236 | 33.5 | 4 | 98.0 | 83 | 2075 | 15.9 | 77 | North America | dodge colt m/m | dodge |
81 | 23.0 | 4 | 120.0 | 97 | 2506 | 14.5 | 72 | Asia | toyouta corona mark ii (sw) | toyota |
118 | 20.0 | 4 | 114.0 | 91 | 2582 | 14.0 | 73 | Europe | audi 100ls | audi |
276 | 31.5 | 4 | 89.0 | 71 | 1990 | 14.9 | 78 | Europe | volkswagen scirocco | volkswagen |
176 | 23.0 | 4 | 120.0 | 88 | 2957 | 17.0 | 75 | Europe | peugeot 504 | peugeot |
autompg_clean.hvplot.hist("weight")
When using by
the plots are overlaid by default. To create subplots instead, use subplots=True
.
autompg_clean.hvplot.hist("weight", by="origin", subplots=True, width=250)
You can also plot histograms of datetime data
import pandas as pd
from bokeh.sampledata.commits import data as commits
commits = commits.reset_index().sort_values("datetime")
commits.head(3)
datetime | day | time | |
---|---|---|---|
4915 | 2012-12-29 11:57:50-06:00 | Sat | 11:57:50 |
4914 | 2013-01-02 17:46:43-06:00 | Wed | 17:46:43 |
4913 | 2013-01-03 16:28:49-06:00 | Thu | 16:28:49 |
commits.hvplot.hist(
"datetime",
bin_range=(pd.Timestamp('2012-11-30'), pd.Timestamp('2017-05-01')),
bins=54,
)
If you want to plot the distribution of a categorical column you can calculate the distribution using Pandas’ method value_counts
and plot it using .hvplot.bar
.
autompg_clean["mfr"].value_counts().hvplot.bar(invert=True, flip_yaxis=True, height=500)