In [1]:

import numpy as np
import pandas as pd
import matplotlib
import matplotlib.pyplot as plt
import seaborn as sns
%matplotlib inline

%load_ext version_information
%load_ext rpy2.ipython

Getting Started With Graphics

When constructing a visualization, we are mapping data elements (annotation, categories, numbers, time series) to visual elements (coordinates, color, size, movement). So the first step in designing a graphic is to decide on

• a data set (usually a DataFrame with observations in rows and variables in columns)
• a mapping (e.g. map height to x-coordinate, weight to y-coordinate, age to color)
• the type of plot(s) desired
  • e.g., bar chart or box plot
  • several types can sometimes be overlaid e.g., rug plot on density plot

After that, we can customize the visual elements in several ways

• direct setting of visual element attributes (size, thickness, color, transparency)
• adding labels (title, subtitle, x-axis label, y-axis label)a
• adding guides (legend, color bar)
• adding annotations (text labels, arrows)
• changing coordinate systems (Cartesian to polar, linear to log)
• changing color scales (color palettes and color maps)
• changing graphic extents (minimum and maximum values displayed)

For global changes to the look and feel of visual elements, we can set styles or themes that simultaneously alter many graphical aspects - background and foreground colors, color scheme, font family used etc.

Sometimes, we need to display multiple plots in a single graphic. To do so, we create a layout that specifies how different plots are related to each other (relative size, sharing of axes). There are two kinds of layouts

• plots are related (i.e. belong to the same data set and type, differ only in choice or subgroup of data elements presented)
• plots are unrelated

Finally, we often need to save the graphic to a file for later viewing or inclusion in a report.

Resources

• Matplotlib
• Seaborn

Data Set

We will load an example data set for the pulse rate after exercise in an experiment with the following variables

• diet
  • low fat
  • no fat
• duration of exercise
  • 1 min
  • 15 min
  • 30 min
• condition
  • rest
  • walking
  • running

In [2]:

exercise = sns.load_dataset("exercise", index_col = 0)

In [3]:

exercise.head()

Out[3]:

	id	diet	pulse	time	kind
0	1	low fat	85	1 min	rest
1	1	low fat	85	15 min	rest
2	1	low fat	88	30 min	rest
3	2	low fat	90	1 min	rest
4	2	low fat	92	15 min	rest

Mapping

Suppose we are interested in the change in pulse after running for 1, 15 or 30 minutes, and also if the type of diet makes a difference. One way to do this is to map time to the x-coordinate, pulse to the y-coordinate and diet to the color. We do this with several types of plots.

In [4]:

df = exercise[exercise.kind == 'running']
df.head()

Out[4]:

	id	diet	pulse	time	kind
60	21	low fat	93	1 min	running
61	21	low fat	98	15 min	running
62	21	low fat	110	30 min	running
63	22	low fat	98	1 min	running
64	22	low fat	104	15 min	running

In [5]:

sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df)
pass

In [6]:

sns.boxplot(x = 'time', y = 'pulse', hue = 'diet', data = df)
pass

Setting visual elements

Change transparency

In [7]:

sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df,
            alpha = 0.5)
pass

Add title

In [8]:

sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df,)
plt.title('Running')
pass

Change coordinate system

In [9]:

sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df)
plt.gca().set(yscale = "log")
pass

Change color palette

In [10]:

sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df,
            palette = 'RdBu_r')
pass

Multiple plots

Suppose we want to see the results for all types of exercise, not just running. We can use multiple plots to achieve this.

In [11]:

sns.factorplot(x = 'time', y = 'pulse', hue = 'diet', col = 'kind',
               kind = 'point', data = exercise)
pass

Global changes with themes

In [12]:

sns.set_context("notebook", font_scale=1.0)
sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df)
pass

In [13]:

sns.set_context("paper", font_scale=2.0)
sns.set_style('white')
sns.barplot(x = 'time', y = 'pulse', hue = 'diet', data = df)
pass

In [14]:

sns.set()
sns.set_context("notebook", font_scale = 1.5)

Saving plots

Preferred

In [15]:

g = sns.factorplot(x = 'time', y = 'pulse', hue = 'diet', col = 'kind',
               kind = 'bar', palette = 'bright', data = exercise)
g.savefig('figs/exercise1.png')

Alternative

In [16]:

sns.factorplot(x = 'time', y = 'pulse', hue = 'diet', col = 'kind',
               kind = 'bar', palette = 'RdBu_r', data = exercise)
plt.savefig('figs/exercise2.png')

Check

In [17]:

from IPython.display import Image

In [18]:

Image('figs/exercise1.png')

Out[18]:

In [19]:

Image('figs/exercise2.png')

Out[19]:

Version Information

In [20]:

%load_ext version_information
%version_information

The version_information extension is already loaded. To reload it, use:
  %reload_ext version_information

Out[20]:

Software	Version
Python	3.5.2 64bit [GCC 4.2.1 Compatible Apple LLVM 4.2 (clang-425.0.28)]
IPython	5.0.0
OS	Darwin 15.6.0 x86_64 i386 64bit
Tue Aug 16 09:04:30 2016 EDT

In [ ]: