Skip to main content

Python Bookdown

15 seaborn

15.1 Seaborn and Matplotlib

  • seaborn returns a matplotlib object that can be modified by the options in the pyplot module
  • Often, these options are wrapped by seaborn and .plot() in pandas and available as arguments

15.2 Sample Data 

n = 100
comp = ['C' + i for i in np.random.randint( 1,4, size  = n).astype(str)] # 3x Company
dept = ['D' + i for i in np.random.randint( 1,4, size  = n).astype(str)] # 5x Department
grp =  ['G' + i for i in np.random.randint( 1,4, size  = n).astype(str)] # 2x Groups
value1 = np.random.normal( loc=50 , scale=5 , size = n)
value2 = np.random.normal( loc=20 , scale=3 , size = n)
value3 = np.random.normal( loc=5 , scale=30 , size = n)

mydf = pd.DataFrame({
    'comp':comp, 
    'dept':dept, 
    'grp': grp,
    'value1':value1, 
    'value2':value2,
    'value3':value3 
})
mydf.head()
#:>   comp dept grp     value1     value2     value3
#:> 0   C3   D2  G3  44.629447  16.439507  48.582119
#:> 1   C2   D2  G2  50.110557  21.988054  34.931015
#:> 2   C3   D2  G1  46.722299  22.605435 -22.025500
#:> 3   C2   D1  G3  58.506386  21.534852  33.198709
#:> 4   C2   D3  G3  40.798755  22.435606 -26.255345

15.3 Scatter Plot

15.3.1 2x Numeric 

sns.lmplot(x='value1', y='value2', data=mydf)
plt.show()
sns.lmplot(x='value1', y='value2', fit_reg=False, data=mydf);  #hide regresion line
plt.show()

15.3.2 2xNumeric + 1x Categorical

Use hue to represent additional categorical feature

sns.lmplot(x='value1', y='value2', data=mydf, hue='comp', fit_reg=False);
plt.show()

15.3.3 2xNumeric + 2x Categorical

Use col and hue to represent two categorical features

sns.lmplot(x='value1', y='value2', col='comp',hue='grp', fit_reg=False, data=mydf);
plt.show()

15.3.4 2xNumeric + 3x Categorical

Use row, col and hue to represent three categorical features

sns.lmplot(x='value1', y='value2', row='dept',col='comp', hue='grp', fit_reg=False, data=mydf);
plt.show()

15.3.5 Customization

15.3.5.1 size

size: height in inch for each facet

sns.lmplot(x='value1', y='value2', col='comp',hue='grp', size=3,fit_reg=False, data=mydf)
plt.show()

Observe that even size is very large, lmplot will fit (shrink) everything into one row by deafult. See example below.

sns.lmplot(x='value1', y='value2', col='comp',hue='grp', size=5,fit_reg=False, data=mydf)
plt.show()

15.3.5.2 col_wrap

To avoid lmplot from shrinking the chart, we use col_wrap=<col_number to wrap the output.
Compare the size (height of each facet) with the above without col_wrap. Below chart is larger.

sns.lmplot(x='value1', y='value2', col='comp',hue='grp', size=5, col_wrap=2, fit_reg=False, data=mydf)
plt.show()

15.4 Histogram 

seaborn.distplot(
  a,               # Series, 1D Array or List
  bins=None,
  hist=True,
  rug = False,
  vertical=False
)

15.4.1 1x Numeric 

sns.distplot(mydf.value1)
plt.show()
sns.distplot(mydf.value1,hist=True,rug=True,vertical=True, bins=30,color='g')
plt.show()

15.5 Bar Chart 

com_grp = mydf.groupby('comp')
grpdf = com_grp['value1'].sum().reset_index()
grpdf
#:>   comp       value1
#:> 0   C1  1777.794043
#:> 1   C2  1834.860416
#:> 2   C3  1343.194018

15.5.1 1x Categorical, 1x Numeric 

sns.barplot(x='comp',y='value1',data=grpdf)
plt.show()

15.5.2 Customization

15.5.2.1 Ordering 

sns.barplot(x='comp',y='value2', hue='grp',
            order=['C3','C2','C1'],
            hue_order=['G1','G2','G3'],
            data=mydf
)
plt.show()

15.5.2.2 Flipping X/Y Axis 

sns.barplot(x='value2',y='comp', hue='grp',data=mydf)
plt.show()

15.6 Faceting

Faceting in Seaborn is a generic function that works with matplotlib various plot utility.
It support matplotlib as well as seaborn plotting utility.

15.6.1 Faceting Histogram 

g = sns.FacetGrid(mydf, col="comp", row='dept')
g.map(plt.hist, "value1")
plt.show()
g = sns.FacetGrid(mydf, col="comp", row='dept')
g.map(plt.hist, "value1")
plt.show()

15.6.2 Faceting Scatter Plot 

g = sns.FacetGrid(mydf, col="comp", row='dept',hue='grp')
g.map(plt.scatter, "value1","value2",alpha=0.7);
g.add_legend()
plt.show()

15.7 Pair Grid

15.7.1 Simple Pair Grid 

g = sns.PairGrid(mydf, hue='comp')
g.map(plt.scatter);
g.add_legend()
plt.show()

15.7.2 Different Diag and OffDiag 

g = sns.PairGrid(mydf, hue='comp')
g.map_diag(plt.hist, bins=15)
g.map_offdiag(plt.scatter)
g.add_legend()
plt.show()