Exploratory Analysis using Pandas, Matplotlib, and Seaborn.¶

In this notebook, we'll start munging some real data and exploring using various Pandas methods and data visualizations.

Imports:

import pandas as pd
import numpy as np
import os
In [1]:
import pandas as pd
import numpy as np
import os

Paths:

dir = r'C:/Users/phwh9568/Workshops/Python_Data_Camp/'
data_dir = os.path.join(dir,'data')
In [2]:
dir = r'C:/Users/phwh9568/Workshops/Python_Data_Camp/'
data_dir = os.path.join(dir,'data')

Let's read in some data from from the EPA's Environmental Justice Screening and Mapping Tool. EJscreen_Colorado.csv is in the data directory. This is an extract of a nationwide dataset available here: https://www.epa.gov/ejscreen/download-ejscreen-data.

We're working with data on the census tract level. You will also find a data dictionary in the data folder that explains the variables: EJScreen_2024_Tract_Percentiles_Columns.xlsx

We'll use this data to make use of Pandas' data munging/manipulating/analyzing capabilities.

Read it in as a variable:

data = pd.read_csv(os.path.join(data_dir,'EJscreen_Colorado.csv'))
In [157]:
data = pd.read_csv(os.path.join(data_dir,'EJScreen_Colorado.csv'),dtype={'ID':str})
In [161]:
data
Out[161]:
ID STATE_NAME ST_ABBREV CNTY_NAME REGION ACSTOTPOP ACSIPOVBAS ACSEDUCBAS ACSTOTHH ACSTOTHU ... PTRAF PRE1960 PRE1960PCT PNPL PRMP PTSDF UST PWDIS NO2 DWATER
0 08001007801 Colorado CO Adams County 8 3889 3850 2338 1288 1403 ... 3.083834e+06 460.0 0.327869 0.669703 1.877369 2.073401 18.367974 6998.315640 7.074942 0.000000
1 08001007802 Colorado CO Adams County 8 4359 4349 2597 1528 1616 ... 2.799872e+06 571.0 0.353342 0.559196 1.580352 2.447664 27.299876 8375.140546 7.646279 0.000000
2 08001007900 Colorado CO Adams County 8 6126 6126 3794 2222 2300 ... 2.896340e+06 1196.0 0.520000 0.692192 2.094691 2.061385 10.328617 8800.094864 7.263358 0.000000
3 08001008000 Colorado CO Adams County 8 5763 5738 3806 1922 2138 ... 2.612349e+06 1295.0 0.605706 0.605078 1.888196 2.437295 6.845090 12517.926928 7.596739 0.000000
4 08001008100 Colorado CO Adams County 8 1674 1477 1305 806 922 ... 2.686544e+06 0.0 0.000000 0.532365 1.639807 3.899799 8.233914 18395.649889 8.273159 0.000000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1442 08123002300 Colorado CO Weld County 8 5470 5432 3940 2168 2233 ... 9.245676e+04 496.0 0.222123 0.000000 0.268646 0.086550 0.106629 607.498975 6.137583 0.370718
1443 08123002501 Colorado CO Weld County 8 5794 5792 4006 1918 2091 ... 2.169755e+04 685.0 0.327594 0.000000 0.330974 0.000000 0.000786 176.948736 3.556240 0.000000
1444 08123002502 Colorado CO Weld County 8 7452 7433 4901 2443 2525 ... 7.159599e+04 602.0 0.238416 0.000000 1.014882 0.159845 0.080032 116.308356 6.791139 0.306212
1445 08125963100 Colorado CO Yuma County 8 4370 4274 2948 1710 1920 ... 4.938223e+04 834.0 0.434375 0.000000 0.665121 0.000000 0.175116 8.536003 5.288863 2.906679
1446 08125963200 Colorado CO Yuma County 8 5568 5378 3718 2243 2393 ... 2.708546e+04 1123.0 0.469285 0.000000 0.214805 0.000000 0.201496 0.006666 6.676397 39.566919

1447 rows × 45 columns

Quick note on field data types... Pandas will guess/assume column data types, and usually this is fine. But not always!

You can explicitly declare a column data type on read using the dtype parameter of .read_csv().

This is me telling you there is a messed up column here. Reimport setting ID to string.

In [162]:
data
Out[162]:
ID STATE_NAME ST_ABBREV CNTY_NAME REGION ACSTOTPOP ACSIPOVBAS ACSEDUCBAS ACSTOTHH ACSTOTHU ... PTRAF PRE1960 PRE1960PCT PNPL PRMP PTSDF UST PWDIS NO2 DWATER
0 08001007801 Colorado CO Adams County 8 3889 3850 2338 1288 1403 ... 3.083834e+06 460.0 0.327869 0.669703 1.877369 2.073401 18.367974 6998.315640 7.074942 0.000000
1 08001007802 Colorado CO Adams County 8 4359 4349 2597 1528 1616 ... 2.799872e+06 571.0 0.353342 0.559196 1.580352 2.447664 27.299876 8375.140546 7.646279 0.000000
2 08001007900 Colorado CO Adams County 8 6126 6126 3794 2222 2300 ... 2.896340e+06 1196.0 0.520000 0.692192 2.094691 2.061385 10.328617 8800.094864 7.263358 0.000000
3 08001008000 Colorado CO Adams County 8 5763 5738 3806 1922 2138 ... 2.612349e+06 1295.0 0.605706 0.605078 1.888196 2.437295 6.845090 12517.926928 7.596739 0.000000
4 08001008100 Colorado CO Adams County 8 1674 1477 1305 806 922 ... 2.686544e+06 0.0 0.000000 0.532365 1.639807 3.899799 8.233914 18395.649889 8.273159 0.000000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1442 08123002300 Colorado CO Weld County 8 5470 5432 3940 2168 2233 ... 9.245676e+04 496.0 0.222123 0.000000 0.268646 0.086550 0.106629 607.498975 6.137583 0.370718
1443 08123002501 Colorado CO Weld County 8 5794 5792 4006 1918 2091 ... 2.169755e+04 685.0 0.327594 0.000000 0.330974 0.000000 0.000786 176.948736 3.556240 0.000000
1444 08123002502 Colorado CO Weld County 8 7452 7433 4901 2443 2525 ... 7.159599e+04 602.0 0.238416 0.000000 1.014882 0.159845 0.080032 116.308356 6.791139 0.306212
1445 08125963100 Colorado CO Yuma County 8 4370 4274 2948 1710 1920 ... 4.938223e+04 834.0 0.434375 0.000000 0.665121 0.000000 0.175116 8.536003 5.288863 2.906679
1446 08125963200 Colorado CO Yuma County 8 5568 5378 3718 2243 2393 ... 2.708546e+04 1123.0 0.469285 0.000000 0.214805 0.000000 0.201496 0.006666 6.676397 39.566919

1447 rows × 45 columns

It's too big to view the whole thing, but we can use various methods to get a feel for the data set...

Let's explore:

data.head()
data.tail()
data.columns
data.describe
In [33]:
data.head()
Out[33]:
ID STATE_NAME ST_ABBREV CNTY_NAME REGION ACSTOTPOP ACSIPOVBAS ACSEDUCBAS ACSTOTHH ACSTOTHU ... PTRAF PRE1960 PRE1960PCT PNPL PRMP PTSDF UST PWDIS NO2 DWATER
0 08001007801 Colorado CO Adams County 8 3889 3850 2338 1288 1403 ... 3.083834e+06 460.0 0.327869 0.669703 1.877369 2.073401 18.367974 6998.315640 7.074942 0.0
1 08001007802 Colorado CO Adams County 8 4359 4349 2597 1528 1616 ... 2.799872e+06 571.0 0.353342 0.559196 1.580352 2.447664 27.299876 8375.140546 7.646279 0.0
2 08001007900 Colorado CO Adams County 8 6126 6126 3794 2222 2300 ... 2.896340e+06 1196.0 0.520000 0.692192 2.094691 2.061385 10.328617 8800.094864 7.263358 0.0
3 08001008000 Colorado CO Adams County 8 5763 5738 3806 1922 2138 ... 2.612349e+06 1295.0 0.605706 0.605078 1.888196 2.437295 6.845090 12517.926928 7.596739 0.0
4 08001008100 Colorado CO Adams County 8 1674 1477 1305 806 922 ... 2.686544e+06 0.0 0.000000 0.532365 1.639807 3.899799 8.233914 18395.649889 8.273159 0.0

5 rows × 45 columns

data.columns
In [159]:
data.columns
Out[159]:
Index(['ID', 'STATE_NAME', 'ST_ABBREV', 'CNTY_NAME', 'REGION', 'ACSTOTPOP',
       'ACSIPOVBAS', 'ACSEDUCBAS', 'ACSTOTHH', 'ACSTOTHU', 'ACSUNEMPBAS',
       'ACSDISABBAS', 'DEMOGIDX_2', 'DEMOGIDX_5', 'PEOPCOLOR', 'PEOPCOLORPCT',
       'LOWINCOME', 'LOWINCPCT', 'UNEMPLOYED', 'UNEMPPCT', 'DISABILITY',
       'DISABILITYPCT', 'LINGISO', 'LINGISOPCT', 'LESSHS', 'LESSHSPCT',
       'UNDER5', 'UNDER5PCT', 'OVER64', 'OVER64PCT', 'LIFEEXPPCT', 'PM25',
       'OZONE', 'DSLPM', 'RSEI_AIR', 'PTRAF', 'PRE1960', 'PRE1960PCT', 'PNPL',
       'PRMP', 'PTSDF', 'UST', 'PWDIS', 'NO2', 'DWATER'],
      dtype='object')

There's a fair amount of data we don't need right now, so let's split off the demographic information we're interested in along with the environmental variables.

Let's make a list of the columns we want:

environmental = list(data.columns[-14:])
In [160]:
environmental = list(data.columns[-14:])
environmental
Out[160]:
['PM25',
 'OZONE',
 'DSLPM',
 'RSEI_AIR',
 'PTRAF',
 'PRE1960',
 'PRE1960PCT',
 'PNPL',
 'PRMP',
 'PTSDF',
 'UST',
 'PWDIS',
 'NO2',
 'DWATER']

And, we'll manually create a list of the demographic variables we want to include:

demographics = ['ID','PEOPCOLORPCT', 'LOWINCPCT', 'LIFEEXPPCT', 'LINGISOPCT', 'DISABILITYPCT']
In [149]:
demographics = ['ID','PEOPCOLORPCT', 'LOWINCPCT', 'LIFEEXPPCT', 'LINGISOPCT', 'DISABILITYPCT']
In [28]:
demographics
Out[28]:
['ID',
 'PEOPCOLORPCT',
 'LOWINCPCT',
 'LIFEEXPPCT',
 'LINGISOPCT',
 'DISABILITYPCT']

Now, combine these two lists then use this list of columns to split off a new dataframe based on the selected columns...

How do we do this?

In [150]:
ejdata = data[demographics + environmental].copy()

Let's join another dataset to the ejscreen data that will tell us if an area is urban or rural.

urban = pd.read_csv(os.path.join(data_dir,'Colorado_Tracts_Urban.csv'), dtype={'GEOID':str})
In [200]:
urban = pd.read_csv(os.path.join(data_dir, 'Colorado_Tracts_Urban.csv'), dtype={'GEOID':str})
In [201]:
urban
Out[201]:
STATEFP COUNTYFP TRACTCE GEOID NAME NAMELSAD MTFCC FUNCSTAT ALAND AWATER INTPTLAT INTPTLON UATYPE20
0 8 27 970102 08027970102 9701.02 Census Tract 9701.02 G5020 S 871112279 455018 38.054321 -105.239722 NaN
1 8 27 970101 08027970101 9701.01 Census Tract 9701.01 G5020 S 1041919696 2909132 38.094958 -105.500103 NaN
2 8 1 8533 08001008533 85.33 Census Tract 85.33 G5020 S 3353029 36543 39.892202 -104.953815 U
3 8 1 8534 08001008534 85.34 Census Tract 85.34 G5020 S 2341349 28063 39.893732 -104.931796 U
4 8 1 8802 08001008802 88.02 Census Tract 88.02 G5020 S 12056169 3131932 39.852058 -104.912748 U
... ... ... ... ... ... ... ... ... ... ... ... ... ...
1442 8 35 14116 08035014116 141.16 Census Tract 141.16 G5020 S 3976441 0 39.549663 -104.877263 U
1443 8 35 14123 08035014123 141.23 Census Tract 141.23 G5020 S 5624145 0 39.424533 -104.881694 U
1444 8 35 13905 08035013905 139.05 Census Tract 139.05 G5020 S 3645627 0 39.513159 -104.756615 U
1445 8 41 6501 08041006501 65.01 Census Tract 65.01 G5020 S 1600530 0 38.800838 -104.737459 U
1446 8 41 2300 08041002300 23.00 Census Tract 23 G5020 S 3129892 0 38.827361 -104.826333 U

1447 rows × 13 columns

Check values of the UATYPE20 column using .unique()

In [204]:
urban['UATYPE20'].unique()
Out[204]:
array(['R', 'U'], dtype=object)

Now, replace NaN values in UATYPE20 to 'R' for rural. Do you remember how?

In [202]:
urban.fillna({'UATYPE20':'R'}, inplace=True)
In [203]:
urban
Out[203]:
STATEFP COUNTYFP TRACTCE GEOID NAME NAMELSAD MTFCC FUNCSTAT ALAND AWATER INTPTLAT INTPTLON UATYPE20
0 8 27 970102 08027970102 9701.02 Census Tract 9701.02 G5020 S 871112279 455018 38.054321 -105.239722 R
1 8 27 970101 08027970101 9701.01 Census Tract 9701.01 G5020 S 1041919696 2909132 38.094958 -105.500103 R
2 8 1 8533 08001008533 85.33 Census Tract 85.33 G5020 S 3353029 36543 39.892202 -104.953815 U
3 8 1 8534 08001008534 85.34 Census Tract 85.34 G5020 S 2341349 28063 39.893732 -104.931796 U
4 8 1 8802 08001008802 88.02 Census Tract 88.02 G5020 S 12056169 3131932 39.852058 -104.912748 U
... ... ... ... ... ... ... ... ... ... ... ... ... ...
1442 8 35 14116 08035014116 141.16 Census Tract 141.16 G5020 S 3976441 0 39.549663 -104.877263 U
1443 8 35 14123 08035014123 141.23 Census Tract 141.23 G5020 S 5624145 0 39.424533 -104.881694 U
1444 8 35 13905 08035013905 139.05 Census Tract 139.05 G5020 S 3645627 0 39.513159 -104.756615 U
1445 8 41 6501 08041006501 65.01 Census Tract 65.01 G5020 S 1600530 0 38.800838 -104.737459 U
1446 8 41 2300 08041002300 23.00 Census Tract 23 G5020 S 3129892 0 38.827361 -104.826333 U

1447 rows × 13 columns

Now merge:

ejdata = ejdata.merge(urban, left_on='ID', right_on='GEOID')
In [198]:
ejdata = ejdata.merge(urban[['GEOID','UATYPE20']], left_on='ID', right_on='GEOID')

Great! Now, export this data to a csv:

ejdata.to_csv(os.path.join(data_dir,'ejdata_urban.csv'))
In [199]:
ejdata.to_csv(os.path.join(data_dir,'ejdata_urban.csv'))

Let's explore this data some. MAYBE we can find some relationships?

Start by using sort values to see census tracts with high ozone levels.

ejdata.sort_values(by=['OZONE'], ascending=False, inplace=True)
In [170]:
ejdata.sort_values(by=['LIFEEXPPCT'], ascending=False, inplace=True)
In [171]:
ejdata
Out[171]:
ID PEOPCOLORPCT LOWINCPCT LIFEEXPPCT LINGISOPCT DISABILITYPCT PM25 OZONE DSLPM RSEI_AIR PTRAF PRE1960 PRE1960PCT PNPL PRMP PTSDF UST PWDIS NO2 DWATER
969 08059011401 0.441455 0.325876 0.309744 0.101921 0.144787 8.590671 77.46888 0.349507 10798.031244 4.052275e+06 404.0 0.331148 0.536284 0.313505 2.612202 9.378742 1084.319116 8.555373 0.094624
75 08001009319 0.643262 0.519858 0.305641 0.127451 0.170213 9.369044 76.84047 0.390803 4865.939296 2.772040e+06 46.0 0.045054 0.649075 0.698815 1.307523 17.556667 4865.903218 9.135113 0.000000
1284 08101002600 0.506517 0.694763 0.303590 0.040323 0.199664 5.939982 69.18587 0.132489 6830.114659 8.323422e+05 480.0 0.299439 0.507024 0.475619 0.573916 5.611099 22474.525460 8.455957 0.000000
1076 08069000802 0.236251 0.352494 0.294359 0.021401 0.083983 7.926807 68.87841 0.181237 9.877347 1.193024e+06 28.0 0.054475 0.000000 0.365951 2.878819 2.990898 4308.660489 12.074300 0.000000
1263 08101000600 0.613076 0.539657 0.294359 0.004802 0.217042 6.063779 69.22637 0.123007 3628.054999 1.354583e+06 791.0 0.765731 0.442512 0.992436 0.522496 7.311510 16000.047701 9.630618 0.000000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1423 08123002017 0.103395 0.041667 NaN 0.000000 0.061728 9.761694 73.85465 0.278583 1572.519479 4.264607e+05 0.0 0.000000 0.000000 0.889166 1.167315 0.000000 455.200607 7.245087 1.000000
1427 08123002021 0.075912 0.110949 NaN 0.000000 0.054332 9.901592 73.34243 0.174785 1294.459181 1.665712e+05 20.0 0.076336 0.000000 0.694326 1.303657 0.000000 1948.832866 5.869461 0.000000
1435 08123002204 0.055610 0.115610 NaN 0.000000 0.126829 9.026354 69.63085 0.191751 68.377096 2.637855e+05 15.0 0.016988 0.000000 0.578050 3.101934 0.899461 16424.483596 7.275004 0.000000
1439 08123002208 0.190395 0.099389 NaN 0.011480 0.081273 8.382512 68.75172 0.117671 57.105426 1.815482e+05 251.0 0.064031 0.000000 0.445016 0.985587 0.050578 13515.802825 6.096531 0.000000
1441 08123002210 0.046149 0.035193 NaN 0.000000 0.046149 9.161547 70.05304 0.206830 39.738557 3.036876e+05 7.0 0.006856 0.000000 0.342369 1.325215 0.000000 12493.865811 6.864555 0.000000

1447 rows × 20 columns

You might find it useful to add some color to your table...

ejdata[['LIFEEXPPCT','OZONE', 'PTRAF', 'PM25']].style.background_gradient()
In [ ]:
ejdata[['LIFEEXPPCT','OZONE', 'PTRAF', 'PM25']].style.background_gradient()

Here's some more info on styling Pandas tables: https://pandas.pydata.org/docs/user_guide/style.html

Let's get into visualization.¶

There's a TON of Python data visualization packages. We'll touch on three big ones:

  1. matplotlib
  2. seaborn
  3. plotly

We'll start with matplotlib:

import matplotlib.pyplot as plt
In [3]:
import matplotlib.pyplot as plt

Let's start by generating some scatterplots.

We'll start with the standard matplotlib approach:

Here are the docs: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.scatter.html.

fig, ax = plt.subplots()

x = ejdata['LIFEEXPPCT']
y = ejdata['PTRAF']

ax.scatter(x,y)

Check the docs and modify... Change the figure size, marker color, marker size, set axis labels.

In [180]:
fig, ax = plt.subplots()

x = ejdata['LIFEEXPPCT']
y = ejdata['PTRAF']

ax.scatter(x,y, c='purple', s=10)
ax.set_xlabel('Proximity to Traffic')
ax.set_ylabel('% Low Life Expectancy')
Out[180]:
Text(0, 0.5, '% Low Life Expectancy')
No description has been provided for this image

Pandas X matplotlib...

Pandas does have some built in matplotlib functionality. For example, https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.plot.scatter.html.

ejdata.plot.scatter(x='LIFEEXPPCT',y='PTRAF')
In [181]:
ejdata.plot.scatter(x='LIFEEXPPCT',
                    y='PTRAF', 
                    xlabel='Perc Low Life Expectancy', 
                    ylabel='Proximity to Traffic', 
                    c = 'magenta',
                    figsize=(8,6))
Out[181]:
<Axes: xlabel='Perc Low Life Expectancy', ylabel='Proximity to Traffic'>
No description has been provided for this image

Definitely easier! But sometimes not as fully customizable.

Although matplotlib is (maybe?) the standard bearer, there are several other very good visualization libraries that can be used as alternatives or alongside matplotlib.

Let's checkout Seaborn¶

import seaborn as sns
In [4]:
import seaborn as sns

Let's start again with our scatterplot.

sns.scatterplot(x=ejdata['LIFEEXPPCT'], Y=EJDATA['PTRAF'])
In [183]:
sns.scatterplot(x=ejdata['LIFEEXPPCT'], y=ejdata['PTRAF'])
Out[183]:
<Axes: xlabel='LIFEEXPPCT', ylabel='PTRAF'>
No description has been provided for this image

Check the docs: Now, recreate the above plot but use the hue parameter to style it according to urban and rural areas:

In [ ]:

Is there something here? Worth exploring? Maybe...

Let's split off just the urban census tracts. Do you remember how to use .loc?

In [60]:
ejUrban = ejdata.loc[ejdata['UATYPE20']=='U'].copy()

I guess we could keep futzing about plugging in various variables into our scatterplots...

But why not produce them all at once?

Let's check out Pairgrid:

g = sns.PairGrid(ejUrban) #or sns.pairplot(ejUrban)
g.map(sns.scatterplot)
In [184]:
g = sns.PairGrid(ejUrban) #or sns.pairplot(ejUrban)
g.map(sns.scatterplot)
Out[184]:
<seaborn.axisgrid.PairGrid at 0x29d168bbfe0>
No description has been provided for this image

Okay, well, we have some unclear results... let's run a correlation:

ejdata.corr(numeric_only=True)
In [185]:
ejdata.corr(numeric_only=True)
Out[185]:
PEOPCOLORPCT LOWINCPCT LIFEEXPPCT LINGISOPCT DISABILITYPCT PM25 OZONE DSLPM RSEI_AIR PTRAF PRE1960 PRE1960PCT PNPL PRMP PTSDF UST PWDIS NO2 DWATER
PEOPCOLORPCT 1.000000 0.513098 0.470926 0.615641 0.185730 0.289236 0.111651 0.344556 -0.001848 0.264043 0.148979 0.183422 0.196012 0.488370 0.252331 0.337754 -0.053133 0.276331 -0.014238
LOWINCPCT 0.513098 1.000000 0.547569 0.401665 0.480656 -0.136454 -0.298828 0.011263 -0.011815 0.062000 0.283148 0.345565 0.084482 0.201260 0.152893 0.324476 0.001464 0.232531 0.010641
LIFEEXPPCT 0.470926 0.547569 1.000000 0.300067 0.341197 0.095128 -0.002337 0.248234 0.069150 0.263584 0.332217 0.371474 0.194439 0.295898 0.255930 0.386097 -0.036149 0.279007 -0.035517
LINGISOPCT 0.615641 0.401665 0.300067 1.000000 0.040846 0.218806 0.057586 0.257970 -0.029970 0.220724 0.119965 0.121186 0.134801 0.320235 0.163974 0.386527 -0.042129 0.166778 -0.020157
DISABILITYPCT 0.185730 0.480656 0.341197 0.040846 1.000000 -0.193762 -0.235499 -0.085096 0.051848 -0.032261 0.176682 0.220600 0.036706 0.043298 -0.066030 0.094810 -0.016352 0.045553 0.031652
PM25 0.289236 -0.136454 0.095128 0.218806 -0.193762 1.000000 0.659994 0.737725 0.091823 0.547214 0.032497 -0.022556 0.171700 0.514393 0.509472 0.319250 0.002245 0.442938 -0.092850
OZONE 0.111651 -0.298828 -0.002337 0.057586 -0.235499 0.659994 1.000000 0.616565 0.175756 0.591289 -0.029476 -0.092660 0.155778 0.245802 0.457902 0.283733 0.073285 0.222540 -0.146583
DSLPM 0.344556 0.011263 0.248234 0.257970 -0.085096 0.737725 0.616565 1.000000 0.103453 0.907728 0.331186 0.275949 0.393211 0.557634 0.653968 0.510720 -0.046036 0.460759 -0.086181
RSEI_AIR -0.001848 -0.011815 0.069150 -0.029970 0.051848 0.091823 0.175756 0.103453 1.000000 0.132017 0.083771 0.081468 0.059624 -0.021502 0.218683 0.136324 0.109132 0.039593 -0.020797
PTRAF 0.264043 0.062000 0.263584 0.220724 -0.032261 0.547214 0.591289 0.907728 0.132017 1.000000 0.405641 0.349043 0.361688 0.431730 0.681706 0.559869 -0.050254 0.383671 -0.080241
PRE1960 0.148979 0.283148 0.332217 0.119965 0.176682 0.032497 -0.029476 0.331186 0.083771 0.405641 1.000000 0.905280 0.322400 0.233661 0.363998 0.337461 -0.035031 0.140414 -0.006913
PRE1960PCT 0.183422 0.345565 0.371474 0.121186 0.220600 -0.022556 -0.092660 0.275949 0.081468 0.349043 0.905280 1.000000 0.320574 0.237122 0.312994 0.315293 -0.010258 0.119117 0.027421
PNPL 0.196012 0.084482 0.194439 0.134801 0.036706 0.171700 0.155778 0.393211 0.059624 0.361688 0.322400 0.320574 1.000000 0.431018 0.390426 0.130028 0.005461 0.106483 0.018529
PRMP 0.488370 0.201260 0.295898 0.320235 0.043298 0.514393 0.245802 0.557634 -0.021502 0.431730 0.233661 0.237122 0.431018 1.000000 0.452939 0.268496 0.076401 0.375085 -0.022703
PTSDF 0.252331 0.152893 0.255930 0.163974 -0.066030 0.509472 0.457902 0.653968 0.218683 0.681706 0.363998 0.312994 0.390426 0.452939 1.000000 0.432191 0.083203 0.377117 -0.070461
UST 0.337754 0.324476 0.386097 0.386527 0.094810 0.319250 0.283733 0.510720 0.136324 0.559869 0.337461 0.315293 0.130028 0.268496 0.432191 1.000000 -0.003644 0.390163 -0.055987
PWDIS -0.053133 0.001464 -0.036149 -0.042129 -0.016352 0.002245 0.073285 -0.046036 0.109132 -0.050254 -0.035031 -0.010258 0.005461 0.076401 0.083203 -0.003644 1.000000 0.043000 -0.019845
NO2 0.276331 0.232531 0.279007 0.166778 0.045553 0.442938 0.222540 0.460759 0.039593 0.383671 0.140414 0.119117 0.106483 0.375085 0.377117 0.390163 0.043000 1.000000 -0.080494
DWATER -0.014238 0.010641 -0.035517 -0.020157 0.031652 -0.092850 -0.146583 -0.086181 -0.020797 -0.080241 -0.006913 0.027421 0.018529 -0.022703 -0.070461 -0.055987 -0.019845 -0.080494 1.000000

Let's visualize that using sns.heatmap().

Check the docs.

In [189]:
sns.heatmap(ejdata.corr(numeric_only=True))
Out[189]:
<Axes: >
No description has been provided for this image

Oh well, I tried! Let's get some different data!

There is a built in dataset in Seaborn:

iris = sns.load_dataset('iris')
In [5]:
iris = sns.load_dataset('iris')

Take a peek...

In [66]:
iris
Out[66]:
sepal_length sepal_width petal_length petal_width species
0 5.1 3.5 1.4 0.2 setosa
1 4.9 3.0 1.4 0.2 setosa
2 4.7 3.2 1.3 0.2 setosa
3 4.6 3.1 1.5 0.2 setosa
4 5.0 3.6 1.4 0.2 setosa
... ... ... ... ... ...
145 6.7 3.0 5.2 2.3 virginica
146 6.3 2.5 5.0 1.9 virginica
147 6.5 3.0 5.2 2.0 virginica
148 6.2 3.4 5.4 2.3 virginica
149 5.9 3.0 5.1 1.8 virginica

150 rows × 5 columns

iris.shape
In [68]:
iris.shape
Out[68]:
(150, 5)
iris.columns
In [65]:
iris.columns
Out[65]:
Index(['sepal_length', 'sepal_width', 'petal_length', 'petal_width',
       'species'],
      dtype='object')

Start with a basic scatterplot using petal length as the x axis and petal width on the y axis.

Once you've got that, color the marker points by species.

In [70]:
sns.scatterplot(x='petal_length', y='petal_width', data=iris, hue='species')
Out[70]:
<Axes: xlabel='petal_length', ylabel='petal_width'>
No description has been provided for this image

Try to make a boxplot!

In [71]:
sns.boxplot(iris)
Out[71]:
<Axes: >
No description has been provided for this image

Make a barplot that shows petal length by species! Make them different colors!

In [91]:
sns.barplot(data=iris, y='species', x = 'petal_length', hue='species')
Out[91]:
<Axes: xlabel='petal_length', ylabel='species'>
No description has been provided for this image

Let's do some more advanced plotting. Say you want to make a figure for your paper...

We'll combine charts using matplotlib. Let's stack these three atop of one another...

Start by reviewing the pyplot.subplots() documentation: https://matplotlib.org/stable/api/_as_gen/matplotlib.pyplot.subplots.html

We'll start with this basic code then modify:

fig, ax = plt.subplots()
In [193]:
fig, ax = plt.subplots()
No description has been provided for this image

Okay, if we want these stacked vertically, we need to set the nrows parameter.
How many rows do we need?

In [194]:
fig, ax = plt.subplots(nrows=3)
No description has been provided for this image

Okay, now we've got 3 empty charts. Let's start by populating the charts.

Review the sns.barplot docs: https://seaborn.pydata.org/generated/seaborn.barplot

What does the ax parameter do?

What happens if we just run the our ax variable?

In [195]:
ax
Out[195]:
array([<Axes: >, <Axes: >, <Axes: >], dtype=object)

Okay, there's 3 things there... how do we select one of them?

In [ ]:

Okay, how can we use this info and apply it to the sns.barplot() ax parameter?

In [196]:
fig, ax = plt.subplots(nrows=3)
sns.barplot(data=iris, x='species', y='petal_length', hue='species', ax=ax[0])
Out[196]:
<Axes: xlabel='species', ylabel='petal_length'>
No description has been provided for this image

Populate the remaining axes with the boxplot and scatterplot.

In [ ]:

Make it pretty.

In [141]:
fig, ax = plt.subplots(nrows=3, figsize=(7,12))
fig.suptitle('My big ass chart')
sns.barplot(data=iris, x='species', y = 'petal_length', hue='species', ax=ax[0])
sns.scatterplot(x='petal_length', y='petal_width', data=iris, hue='species', ax=ax[1])
sns.boxplot(iris)
Out[141]:
<Axes: >
No description has been provided for this image