Seaborn for Data Visualization

Based on a tutorial by Shahidhya Ramachandran

https://seaborn.pydata.org/tutorial.html ## Check out matplotlib plots
Method for saving plots to disk (especially useful for including them in your reports:
https://matplotlib.org/devdocs/api/_as_gen/matplotlib.pyplot.savefig.html

Import Packages: Seaborn, Pandas, Matplotlib

iPython magic coe %matplotlib inline allows for inline display of plots
normally you have to do plt.show() to display your plots.



In [14]:

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

Load the dataset: low_birth_rate_full.csv

After loading, explore the data, make sure everything loaded properly
Relabel headers: overwrite them first thing after reading the data.

Explore data frames to check headers and data types

low: indicator of birth weight less than 2.5kg
age: mother's age in years
lwt: mother's weight in pounds at last menstrual period
race: mothers race ("white", "black", "other")
smoke: smoking status during pregnancy
ht: history of hypertension
ui: presence of uterine irritability
ftv: number of physician visits during the first trimester
ptl: number of previous premature labours
bwt: birth weight in grams



In [3]:

    
df = pd.read_csv("~/low_birth_rate_full.csv")
df.columns = ["ID","LOW","AGE","LWT","RACE","SMOKE","PTL","HT","UI","FTV","BWT"]
df.info()









    



<class 'pandas.core.frame.DataFrame'>
RangeIndex: 189 entries, 0 to 188
Data columns (total 11 columns):
ID       189 non-null int64
LOW      189 non-null int64
AGE      189 non-null int64
LWT      189 non-null int64
RACE     189 non-null int64
SMOKE    189 non-null int64
PTL      189 non-null int64
HT       189 non-null int64
UI       189 non-null int64
FTV      189 non-null int64
BWT      189 non-null int64
dtypes: int64(11)
memory usage: 16.3 KB

First plot: scatterplot with linear correlation

Compare birth weight and age.

Older mothers have a harder time giving birth, so maybe there's a relationship there.
Pass BWT as Y variable and AGE as X variable to seaborns lmplot method (linear model plot)
It plot the points, axes, and regression line, and also plots an error field. Super handy!



In [4]:

    
sns.set(style="ticks")
sns.lmplot(y='BWT',x='AGE',data=df)









    Out[4]:





<seaborn.axisgrid.FacetGrid at 0x1135b6860>

Check how weight of mother effects child, controlling for race.

No real hypothes, just to show you how we can do this.
Code for race: 1=white, 2=black, 3=other
Use command below to save this plot



In [5]:

    
sns.lmplot(y="BWT",x="LWT",hue="RACE",data=df) 

p = sns.lmplot(y="BWT",x="LWT",hue="RACE",data=df) 
p.savefig("fancy-regression-chart.png")

Second plot: Violin Plots

Violin plots are an excellent way to explore data.
Similar to visual t-tests, evaluate plots to guage difference (instead of p-value and R-squared)
Does smoking status of mother impact birthweight of baby?



In [6]:

    
sns.violinplot(y="BWT",x="SMOKE",data=df)









    Out[6]:





<matplotlib.axes._subplots.AxesSubplot at 0x116c99ac8>



In [ ]:

    
### Plot 'SMOKE' status, infant 'BWT', and hyptertension 'HT'
* Qualify difference between smoking and non-smoking mothers, with hypertension 
* What happens when we add hypertension into the mix?



In [7]:

    
sns.violinplot(y="BWT",x="HT",hue="SMOKE",data=df)









    Out[7]:





<matplotlib.axes._subplots.AxesSubplot at 0x116eabb38>

Third Plot: Factorplot

Compare interaction of SMOKE, BWT, HT using bar charts.



In [8]:

    
sns.factorplot(x="SMOKE",hue="LOW",col="HT",kind="count",data=df)









    Out[8]:





<seaborn.axisgrid.FacetGrid at 0x117185eb8>

Fourth plot: Correlation Heat Map

Import numberical python package NumPy

Analyze correlation between all pairs of variables in heatmap

Generate a mask for the upper triangle
Set up matplotlib figure
Generate a custom diverging colormap
Draw heatmap with the mask and correct aspect ratio



In [11]:

    
import numpy as np
corr = df.corr()

mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True

f, ax = plt.subplots(figsize=(11, 9))

cmap = sns.diverging_palette(220, 10, as_cmap=True)

sns.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})









    Out[11]:





<matplotlib.axes._subplots.AxesSubplot at 0x116b75c18>

Fifth Plot: Pairplots

To understand the distribution of each variable and
Also plot it against all other variables to understand their relationship.
Graph can be visualized for different values of a chosen 'hue' variable



In [12]:

    
df1 = df[['AGE','LWT','BWT','FTV','PTL','LOW']]
sns.pairplot(df1, hue = 'LOW',size=2.5);

Sixth Plot: Histogram with Kernel Density Estimate

To understand the distribbution of a single variable
Plot the histogram and overlay the distribution of the variable.
Helps identify skew in the data.



In [13]:

    
sns.distplot(df['LWT'], bins = 20, rug = True)









    Out[13]:





<matplotlib.axes._subplots.AxesSubplot at 0x119406c18>



In [ ]: