Plotting continuous outcome and its related variables using seaborn and matplotlib

References, resources and libraries

matplotlib: https://matplotlib.org/

seaborn: https://seaborn.pydata.org/

seaborn axis_grids: https://seaborn.pydata.org/tutorial/axis_grids.html

library(reticulate)

import sys

print(sys.version)

## 3.7.5 (default, Oct 31 2019, 15:18:51) [MSC v.1916 64 bit (AMD64)]

import sys

if not sys.warnoptions:
    import warnings
    warnings.simplefilter("ignore")

import numpy as np
import pandas as pd
from pandas import Series, DataFrame

import matplotlib.pyplot as plt

import matplotlib as mpl

mpl.use('ps') # generate postscript output by default

import seaborn as sb

sb.set_style('whitegrid')

import sklearn
from sklearn import preprocessing
from sklearn.preprocessing import scale

pd.set_option('precision', 3)

pd.set_option('expand_frame_repr', True)

#pd.set_option('max_colwidth', -1)

Structure of the dataset

library(MASS)

data("birthwt")

str(birthwt)

## 'data.frame':    189 obs. of  10 variables:
##  $ low  : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ age  : int  19 33 20 21 18 21 22 17 29 26 ...
##  $ lwt  : int  182 155 105 108 107 124 118 103 123 113 ...
##  $ race : int  2 3 1 1 1 3 1 3 1 1 ...
##  $ smoke: int  0 0 1 1 1 0 0 0 1 1 ...
##  $ ptl  : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ ht   : int  0 0 0 0 0 0 0 0 0 0 ...
##  $ ui   : int  1 0 0 1 1 0 0 0 0 0 ...
##  $ ftv  : int  0 3 1 2 0 0 1 1 1 0 ...
##  $ bwt  : int  2523 2551 2557 2594 2600 2622 2637 2637 2663 2665 ...

Describe the variables

#	variable name	variable label	coded levels
1	low	indicator of birth weight less than 2.5 kg	0, 1
2	age	mother’s age in years	continous variable
3	lwt	mother’s weight in pounds at last menstrual period	continous variable
4	race	mother’s race (1 = white, 2 = black, 3 = other)	1, 2, 3
5	smoke	smoking status during pregnancy	0, 1
6	ptl	number of previous premature labours	0, 1, 2, 3
7	ht	history of hypertension	0, 1
8	ui	presence of uterine irritability	0, 1
9	ftv	number of physician visits during the first trimester	0, 1, 2, 3, 4, 6
10	bwt	birth weight in grams	continous variable

The continuous outcome variable is bwt. Its related variables are: age, lwt, race, smoke, ptl, ht, ui and ftv.

Import and glimpse the data

# save the 'birthwt' data as a 'csv' file and import using pandas 'read_csv' function 

pbwt=pd.read_csv('birthwt.csv')

Information about the data set

# information about data set
pbwt.info()

## <class 'pandas.core.frame.DataFrame'>
## RangeIndex: 189 entries, 0 to 188
## Data columns (total 10 columns):
## 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(10)
## memory usage: 14.9 KB

Data types

# data types: integer/object/category/floating-point

pbwt.dtypes

## low      int64
## age      int64
## lwt      int64
## race     int64
## smoke    int64
## ptl      int64
## ht       int64
## ui       int64
## ftv      int64
## bwt      int64
## dtype: object

First five rows of data

pbwt.head()

##    low  age  lwt  race  smoke  ptl  ht  ui  ftv   bwt
## 0    0   19  182     2      0    0   0   1    0  2523
## 1    0   33  155     3      0    0   0   0    3  2551
## 2    0   20  105     1      1    0   0   0    1  2557
## 3    0   21  108     1      1    0   0   1    2  2594
## 4    0   18  107     1      1    0   0   1    0  2600

Last five rows of data

pbwt.tail()

##      low  age  lwt  race  smoke  ptl  ht  ui  ftv   bwt
## 184    1   28   95     1      1    0   0   0    2  2466
## 185    1   14  100     3      0    0   0   0    2  2495
## 186    1   23   94     3      1    0   0   0    0  2495
## 187    1   17  142     2      0    0   1   0    0  2495
## 188    1   21  130     1      1    0   1   0    3  2495

Data shape or dimension

pbwt.shape

## (189, 10)

Columns name

pbwt.columns

## Index(['low', 'age', 'lwt', 'race', 'smoke', 'ptl', 'ht', 'ui', 'ftv', 'bwt'], dtype='object')

Number of columns

len(pbwt.columns)

## 10

Number of observations

len(pbwt)

## 189

Length of a variable

len(pbwt['age'])

## 189

Number of rows

# rows

pbwt.index

## RangeIndex(start=0, stop=189, step=1)

Plotting continuous outcome and its related varables using seaborn

Distplot: Distribution (histogram, kernel density and rug) plot

# only histogram

#f, ax = plt.subplots(figsize=(8, 4))

sb.distplot(pbwt['bwt'].dropna(), kde=False, bins=13, color='red')

plt.title('Distribution of infant body weight')

plt.show()

# only kernel density plot

sb.distplot(pbwt['bwt'], hist=False, kde=True, rug=False)

plt.title('KDE plot of infant body weight')

plt.show()

# histogram with kernel density plot

sb.distplot(pbwt['bwt'])

plt.title('Distribution of infant body weight')

plt.show()

# histogram with kernel density and rug plot

sb.distplot(pbwt['bwt'], rug=True)

plt.title('Distribution of infant body weight')

plt.show()

Relplot: Scatter plots

# Scatter plot of lwt and bwt

g=sb.relplot(x="lwt", y="bwt", data=pbwt, height=8, aspect=1)

plt.title('Scatter plot of lwt and bwt')

plt.show(g)

# Scatter plot of lwt and bwt in relation to smoking status

g=sb.relplot(x="lwt", y="bwt", hue='smoke', data=pbwt, height=8)

plt.title("Scatter plot of lwt and bwt in relation to smoking status")

plt.show(g)

#Scatter plot of lwt and bwt in relation to smoking status: scatter points denoted in different styles

g=sb.relplot(x="lwt", y="bwt", hue='smoke', style='smoke', data=pbwt, height=8)

plt.title("Scatter plot of lwt and bwt in relation to smoking status")

plt.show(g)

# Scatter plot of lwt and bwt in relation to smoking status: scatter points denoted in different size

g=sb.relplot(x="lwt", y="bwt", hue='smoke', size='smoke', sizes=(15, 150), data=pbwt, height=8)

plt.title("Scatter plot of lwt and bwt in relation to smoking status")

plt.show(g)

# Scatter plot of lwt and bwt in relation to smoking status in different racial groups

g=sb.relplot(x="lwt", y="bwt", hue='smoke', size='smoke', col='race', data=pbwt, height=8)

plt.show(g)

Catplot: Categorical scatter plot

#Distribution of bwt in relation to race: Without jitter

g=sb.catplot(x="race", y="bwt", jitter= False, data=pbwt, height=8)

plt.title("Distribution of bwt in relation to race")

plt.show(g)

# Distribution of bwt in relation to race: With jitter

g=sb.catplot(x="race", y="bwt", data=pbwt, height=8)

plt.title("Distribution of bwt in relation to race")

plt.show(g)

# Distribution of bwt in relation to race: Swarm plot and ordering the categorical groups according to your choice

g=sb.catplot(x="race", y="bwt", kind='swarm', data=pbwt, order=[2, 3, 1], height=8)

plt.title("Distribution of bwt in relation to race")

plt.show(g)

# for selected categories: omit one group from plotting

g=sb.catplot(x="race", y="bwt", kind="swarm", data=pbwt.query("race != 3"), height=8)

plt.title("Distribution of bwt in relation to race")

plt.show(g)

# for selected categories of a 3rd variable (e.g., smoker)

g=sb.catplot(x="race", y="bwt", kind="swarm", data=pbwt.query("smoke == 0"), height=8)

plt.title("Distribution of bwt in non-smokers in relation to race ")

plt.show(g)

# Distribution of bwt in relation to race in smoking categories

g = sb.catplot(x="race", y="bwt", col="smoke", aspect=1,  kind="swarm", data=pbwt, height=8)

plt.show(g)

Pairplot: Pair plots

For a dataframe

# Pair plot for all variables

g = sb.pairplot(pbwt)

plt.show(g)

# alternatively using seaborn PairGrid

g = sb.PairGrid(pbwt)

g = g.map(plt.scatter)

plt.show(g)

# using seaborn PairGrid but also mapping to diagonal and off-diagonal grid

g = sb.PairGrid(pbwt)

g = g.map_diag(plt.hist)

g = g.map_offdiag(plt.scatter)

plt.show(g)

For a subset of variables in the dataframe

sb.pairplot(pbwt[['age', 'lwt', 'bwt']], height=3)

## <seaborn.axisgrid.PairGrid object at 0x0000000037F0F488>

plt.show()

For a subset of variables categorized by the levels of another variable

g = sb.pairplot(pbwt[['age', 'lwt', 'bwt', 'race']], hue='race', palette='hls', height=3)

plt.show(g)

# For a subset of continuous variables grouped by another categorical variable: with labels of categorical groups

g = sb.pairplot(pbwt[['age', 'lwt', 'bwt', 'race']].replace({'race': {1:'white', 2:'black', 3:'other'}}), hue='race', palette='hls', height=3)

plt.show(g)

With regression line

# pair plot with regression line

g = sb.pairplot(data=pbwt, x_vars=["age", "lwt"], y_vars=["bwt"], aspect=1, kind="reg", height=8)

plt.show(g)

# pair plot with regression line: scatter points identified as a categorical variable (e.g., race)

g = sb.pairplot(data=pbwt, x_vars=["age", "lwt"], y_vars=["bwt"], hue='race', aspect=1, kind="reg", height=8)

plt.show(g)

Regplot: Regression plot

# baseline regression line

sb.regplot(x='lwt', y='bwt', data=pbwt, scatter=False, color='red')

plt.title("Baseline regression plot for lwt abd bwt")

plt.show()

# regression plot with scatters

sb.regplot(x='lwt', y='bwt', data=pbwt, scatter=True, color='red')

plt.title("Regression plot for lwt and bwt")

plt.show()

# regression plot with resizing ability: using add_axes

fig= plt.figure(figsize=(8,8))

ax= fig.add_axes([0.1,0.1,0.8,0.8])

sb.regplot(x='lwt', y='bwt', data=pbwt, scatter=True, color='red', ax=ax)

ax.set_title("Regression plot for lwt and bwt")

plt.show()

# regression plot with resizing ability: using plt subplots

fig, ax = plt.subplots(figsize=(8, 8))

sb.regplot(x='lwt', y='bwt', data=pbwt, scatter=True, color='red', ax=ax)

#plt.title("Regression plot for lwt and bwt")

ax.set_title("Regression plot for lwt and bwt")

plt.show()

Residplot: Residual plot

# residual plot 

f, ax = plt.subplots(figsize=(8, 8))

sb.residplot(x="lwt", y="bwt", data=pbwt, scatter_kws={"s": 80}, color='red', ax=ax)

plt.title("Residual plot for lwt and bwt")

plt.show()

Regression and residual plots together

# regression plot with resizing ability

fig, axes= plt.subplots(nrows=1, ncols=2, figsize=(16,8))


# regression plot 

sb.regplot(x='lwt', y='bwt', data=pbwt, scatter=True, scatter_kws={"s": 50}, color='red', ax=axes[0])

axes[0].set_title("Regression plot for lwt and bwt")



# residual plot 

sb.residplot(x="lwt", y="bwt", data=pbwt, scatter_kws={"s": 50}, color='black', ax=axes[1])

axes[1].set_title("Residual plot for lwt and bwt")


#plt.tight_layout()

plt.show()

Lmplot: Linear model plot

With linear fit

# lm plot with different markers and colors

g = sb.lmplot(x='lwt', y='bwt', hue='race', data=pbwt, markers=["o", "x", "^"], palette="Set1", aspect=1, height=8)

plt.title("Linear relationship between lwt and bwt in 3 racial groups")

plt.show(g)

# lm plot for lwt and bwt for different racial groups with different smoking status

g = sb.lmplot(x='lwt', y='bwt', hue='race', col='smoke', data=pbwt, aspect=1, col_wrap=2, height=8)

plt.show(g)

# lm plot for lwt and bwt for individuals with different smoking status within diffrent (racial X ui) groups")

g = sb.lmplot(x='lwt', y='bwt', hue='smoke',  col='race', row='ui', data=pbwt)

plt.show(g)

# lm plot for lwt and bwt for smokers within diffrent (racial X ui) groups") for the levels of another categorical variable

g = sb.lmplot(x='lwt', y='bwt', hue='smoke', col='race', row='ui', data=pbwt.query("ftv==0"), ci=None, scatter_kws={"s": 80})

plt.show(g)

With quadratic fit

# lm plot with quadratic fit

g = sb.lmplot(x='lwt', y='bwt', hue='smoke',  col='race', row='ui', data=pbwt.query("ftv==0"), order=2, ci=None, scatter_kws={"s": 80})

plt.show(g)

With lowess fit

# lm plot with lowess fit

g = sb.lmplot(x="lwt", y="bwt", data=pbwt, lowess=True, height=8)

plt.title("Lowess plot for lwt and bwt")

plt.show(g)

With robust regression fit

# Robust regression between lwt and bwt

g = sb.lmplot(x="age", y="bwt", data=pbwt.query('smoke==0'), robust=True, height=8, ci=None, scatter_kws={"s": 80})

plt.title("Robust regression between lwt and bwt")

plt.show(g)

Boxplots: Using boxplot()

pbwt.dtypes

## low      int64
## age      int64
## lwt      int64
## race     int64
## smoke    int64
## ptl      int64
## ht       int64
## ui       int64
## ftv      int64
## bwt      int64
## dtype: object

For a continuous variable

# box plot for bwt

fig, ax = plt.subplots(figsize=(6, 4))

sb.boxplot(x='bwt', data=pbwt, ax=ax)

plt.title("Distribution of bwt")

plt.show()

For a continuous variable categorized by the levels of another variable

# box plot for bwt by race: axis would determine orientation

sb.boxplot(x='bwt', y=pbwt['race'].astype('category'), data=pbwt, palette='hls')

plt.title("Distribution of bwt by race")

plt.show()

Boxplots: Using catplot()

For a continuous variable

# box plot for bwt by race

g = sb.catplot(x="bwt", kind="box", data=pbwt)

plt.title("Distribution of bwt by race")

plt.show(g)

For a subset of continuous variables

# horizontal box plot using a subset of the data frame or selected variables

g = sb.catplot(data=pbwt[['lwt', 'bwt']], orient='h', kind="box", height=4, aspect=2)

plt.title("Distribution of lwt and bwt")

plt.show(g)

For a continuous variable grouped by another categorical variable

# box plot for bwt by race

pbwt['race_c']=pbwt['race'].astype('category')

g = sb.catplot(y='race_c', x='bwt', kind="box", data=pbwt, height=8)

plt.title("Distribution of bwt by race")

plt.show(g)

For a continuous variable grouped by two categorical variables

# box plot for bwt by (raceXsmoke) groups

pbwt['race_c']=pbwt['race'].astype('category')

g = sb.catplot(x="bwt", y="race_c", hue='smoke', kind="box", data=pbwt, height=8)

plt.title("Distribution of bwt by smoking status in different racial groups")

plt.show(g)

For selected subgroups

# box plot for selected subgroups

g = sb.catplot(y="race", x="bwt", row="smoke", kind="box", orient="h", height=3, aspect=4, data=pbwt.query('race==1'))

plt.show(g)

# box plot for selected subgroups

g = sb.catplot(x="bwt", y="race", row="smoke", kind="box", orient="h", height=3, aspect=4, data=pbwt.query('race!=1'))

plt.show(g)

Plotting continuous outcome and its related variables using matplotlib

Line charts using plot.line(), plot(kind='line') or plt.plot()

pbwt[‘bwt’].plot.hist()

fig, ax = plt.subplots(figsize=(8, 8))


pbwt['bwt'].plot.line(color = 'red')

#pbwt['bwt'].plot(color = 'red')

#pbwt['bwt'].plot(kind='line', color = 'red')

plt.xlabel("Observation ID")

plt.ylabel("bwt (grams)")

plt.title("Variation in bwt")

plt.show()

Line chart with maximum value of a variable identified

# what is the maximum value of bwt

pbwt.bwt.max()

## 4990

# what is the observation/row number that contains the maximum value of bwt

pbwt.query('bwt==4990')

# pbwt[pbwt['bwt']==4990]

##      low  age  lwt  race  smoke  ptl  ht  ui  ftv   bwt race_c
## 129    0   45  123     1      0    0   0   0    1  4990      1

## object oriented programming

fig, ax = plt.subplots(figsize=(8, 8))

ax.plot(pbwt['bwt'], color='red')

#plt.plot(pbwt['bwt'], color='red')

ax.set_xlabel('Obs ID')

ax.set_ylabel('bwt')

ax.set_title('Distribution of bwt')

ax.annotate('Highest bwt', xy= (130, 5000), xytext=(140, 5100), arrowprops=dict(facecolor='black', shrink=0.05))

plt.show()

Line chart scaled using scale() function

#standardized_bwt = scale(pbwt['bwt'], axis=0, with_mean=False, with_std=False)

standardized_bwt = scale(pbwt['bwt'])

plt.plot(standardized_bwt)

plt.title('Scaled or standardized bwt')

plt.show()

Line chart scaled using preprocessing.MinMaxScaler() function

pbwt['bwt'].values.reshape(-1, 1)[0:5]

## array([[2523],
##        [2551],
##        [2557],
##        [2594],
##        [2600]], dtype=int64)

scaled = preprocessing.MinMaxScaler()

bwt_matrix = pbwt['bwt'].values.reshape(-1,1)

scaled_bwt = scaled.fit_transform(bwt_matrix)

plt.plot(scaled_bwt)

plt.title("Scaled or standardized line chart")

plt.show()

Line chart scaled to a range using preprocessing.MinMaxScaler(feature_range=...) function

scaled = preprocessing.MinMaxScaler(feature_range=(15, 20))


bwt_matrix = pbwt['bwt'].values.reshape(-1,1)

scaled_bwt = scaled.fit_transform(bwt_matrix)

plt.plot(scaled_bwt)

plt.title("Line chart scaled or standardized to a range")

plt.show()

Line chart for more than one variable

# put both plots together

fig = plt.figure()

fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2, figsize=(16,8))


ax1.plot(pbwt['bwt'], color = 'blue')

ax1.set_xlabel('Obs ID')

ax1.set_ylabel('bwt')

ax1.set_title('Distribution of bwt')

ax1.annotate('Highest bwt', xy= (130, 5000), xytext=(150, 5100), arrowprops=dict(facecolor='black', shrink=0.05))


ax2.plot(pbwt['lwt'], color = 'red')

ax2.set_xlabel('Obs ID')

ax2.set_ylabel('lwt')

ax2.set_title('Distribution of lwt')

ax2.annotate('Highest lwt', xy= (68, 250), xytext=(90, 253), arrowprops=dict(facecolor='black', shrink=0.05))

plt.show()

Histograms using plot.hist(), plot(kind='hist') or plt.hist()

pbwt['bwt'].plot.hist()

#pbwt['bwt'].plot(kind='hist')

# plt.hist(pbwt['bwt'])

plt.title('Histogram for bwt')

plt.show()

Scatter plots using plot.scatter(), plot(kind='scatter') or plt.scatter()

pbwt.plot.scatter(x='lwt', y='bwt', c=['red'], s=25)

#pbwt.plot(kind='scatter', x='lwt', y='bwt', c=['red'], s=25)

#plt.scatter(x='lwt', y='bwt', c=['red'], s=25, data=pbwt)

#plt.scatter(pbwt['lwt'], pbwt['bwt'], c=['red'], s=25)

#plt.scatter(x=pbwt['lwt'], y=pbwt['bwt'], c=['red'], s=25)

plt.title('Scatter plots for lwt and bwt')

plt.show()

Box plots using plot.box(), boxplot() or plt.boxplot()

For a variable

pbwt['bwt'].plot.box()

#pbwt['bwt'].plot(kind='box')

#pbwt[['bwt']].boxplot()

#pbwt.boxplot(column= 'bwt')

#plt.boxplot(pbwt['bwt'])

plt.title("Distribution of bwt")

plt.show()

For multiple variables

pbwt[['age', 'lwt', 'bwt']].plot.box()

#pbwt[['age', 'lwt', 'bwt']].boxplot()

plt.title("Distribution of age, lwt and bwt")

plt.show()

For a continuous variable grouped by a categorical variable

pbwt.boxplot(column='bwt', by='race')

#plt.title("Distribution of bwt by race")

plt.show()