Simple plotting library for both long and wide data integrated with DataFrames
Project description
Dexplot
A Python library for making data visualizations.
The current aim of Dexplot is to make data visualization creation in Python more robust and straightforward. Dexplot is built on top of Matplotlib and accepts Pandas DataFrames as inputs.
Installation
pip install dexplot
Goals
The primary goals for Dexplot are:
- Maintain a very consistent API with as few functions as necessary to make the desired statistical plots
- Allow the user to tweak the plots without digging into Matplotlib
Tidy Data from Pandas
Dexplot only accepts Pandas DataFrames as input for its plotting functions that are in "tidy" form.
Sample plots
Dexplot currently maintains one primary function, aggplot
which is used to aggregate data and can create five different kinds of plots.
bar
line
box
hist
kde
There are 7 primary parameters to aggplot
:
agg
- Name of column to be aggregated. If it is a column with string/categorical values, then the counts or frequency percentage will be returned.groupby
- Name of column whose unique values will form independent groups. This is used in a similar fashion as thegroup by
SQL clause.data
- The Pandas DataFramehue
- The name of the column to further group the data within a single plotrow
- The name of the column who's unique values split the data in to separate rowscol
- The name of the column who's unique values split the data in to separate columnskind
- The kind of plot to create. One of the five strings from above.
City of Houston Data
To get started, we will use City of Houston employee data collected from the year 2016. It contains public information from about 1500 employees and is located in Dexplot's GitHub repository.
import pandas as pd
import dexplot as dxp
emp = pd.read_csv('data/employee.csv')
emp.head()
title | dept | salary | race | gender | experience | experience_level | |
---|---|---|---|---|---|---|---|
0 | POLICE OFFICER | Houston Police Department-HPD | 45279.0 | White | Male | 1 | Novice |
1 | ENGINEER/OPERATOR | Houston Fire Department (HFD) | 63166.0 | White | Male | 34 | Veteran |
2 | SENIOR POLICE OFFICER | Houston Police Department-HPD | 66614.0 | Black | Male | 32 | Veteran |
3 | ENGINEER | Public Works & Engineering-PWE | 71680.0 | Asian | Male | 4 | Novice |
4 | CARPENTER | Houston Airport System (HAS) | 42390.0 | White | Male | 3 | Novice |
Plotting the average salary by department
The agg
parameter is very important and is what will be aggregated (summarized by a single point statistic, like the mean or median). It is the first parameter and only parameter you must specify (besides data
). If this column is numeric, then by default, the mean of it will be calculated. Here, we specify the groupby
parameter, who's unique values form the independent groups and label the x-axis.
dxp.aggplot(agg='salary', groupby='dept', data=emp)
<matplotlib.axes._subplots.AxesSubplot at 0x1146b7550>
Make horizontal with the orient
parameter
The orient
parameter controls whether the plot will be horizontal or vertical. By default it is set to 'h'
.
dxp.aggplot(agg='salary', groupby='dept', data=emp, orient='h')
<matplotlib.axes._subplots.AxesSubplot at 0x114a37438>
Controlling the figure size
One of the goals of Dexplot is to not have you dip down into the details of Matplotlib. We can use the figsize
parameter to change the size of our plot.
dxp.aggplot(agg='salary', groupby='dept', data=emp, orient='h', figsize=(8, 4))
<matplotlib.axes._subplots.AxesSubplot at 0x1149cee80>
Adding another dimension with hue
The hue
parameter may be used to further subdivide each unique value in the groupby
column. Notice that long tick labels are automatically wrapped.
dxp.aggplot(agg='salary', groupby='dept', data=emp, hue='gender')
<matplotlib.axes._subplots.AxesSubplot at 0x1170f2518>
Aggregating a String/Categorical column
It is possible to use a string/categorical column as the aggregating variable. In this instance, the counts of the unique values of that column will be returned. Because this is already doing a groupby
, you cannot specify a groupby
column in this instance. Let's get the count of employees by race.
dxp.aggplot(agg='race', data=emp, figsize=(8, 4))
<matplotlib.axes._subplots.AxesSubplot at 0x1173e1fd0>
Using hue
with a String/Categorical column
Using a groupby
is not allowed when a string/categorical column is being aggregated. But, we can still sub-divide the groups further by specifying hue
.
dxp.aggplot(agg='race', data=emp, hue='dept')
<matplotlib.axes._subplots.AxesSubplot at 0x1176cf6d8>
Getting the frequency percentage with normalize
It is possible to turn the raw counts into percentages by passing a value to normalize
. Let's find the percentage of all employees by race.
dxp.aggplot(agg='race', data=emp, normalize='all', figsize=(8, 4))
<matplotlib.axes._subplots.AxesSubplot at 0x1171bdba8>
You can normalize over any variable
The parameter normalize
can be either 'agg'
, 'hue'
, 'row'
, 'col'
, or a tuple containing any number of these or 'all'
. For instance, in the following plot, you can normalize by either agg
or hue
.
dxp.aggplot(agg='race', data=emp, hue='dept', normalize='agg')
<matplotlib.axes._subplots.AxesSubplot at 0x117abc6a0>
Data normalized by race
As you can see, the data was normalized by race. For example, from the graph, we can tell that about 30% of black employees were members of the police department. We can also normalize by department. From the graph, about 10% of the Health & Human Services employees were Asian.
dxp.aggplot(agg='race', data=emp, hue='dept', normalize='hue')
<matplotlib.axes._subplots.AxesSubplot at 0x117c0fb38>
Other kinds of plots line
, box
, hist
, and kde
aggplot
is capable of making four other kinds of plots. The line
plot is very similar to the bar plot but simply connects the values together. Let's go back to a numeric column and calculate the median salary by department across each gender.
dxp.aggplot(agg='salary', data=emp, groupby='dept', hue='gender', kind='line', aggfunc='median')
<matplotlib.axes._subplots.AxesSubplot at 0x117fe55f8>
aggfunc
can take any string value that Pandas can
There are more than a dozen string values that aggfunc
can take. These are simply passed to Pandas groupby
method which does the aggregation.
All plots can be both vertical and horizontal
We can rotate all plots with orient
.
dxp.aggplot(agg='salary', data=emp, groupby='dept', hue='gender', kind='line', aggfunc='median', orient='h')
<matplotlib.axes._subplots.AxesSubplot at 0x1181b6240>
Boxplots
Here is the same data plotted as a box plot. This isn't actually an aggregation, so the aggfunc
parameter is meaningless here. Instead, all the values of the particular group are plotted.
dxp.aggplot(agg='salary', data=emp, groupby='dept', hue='gender', kind='box', orient='h')
<matplotlib.axes._subplots.AxesSubplot at 0x118379390>
Histograms and KDE's
As with boxplots, histograms and kdes do not function with aggfunc
as they aren't aggregating but simply displaying all the data for us. Also, it is not possible to use both groupby
and agg
with these plots.
dxp.aggplot(agg='salary', data=emp, groupby='dept', kind='hist', orient='v')
<matplotlib.axes._subplots.AxesSubplot at 0x118a7ac88>
dxp.aggplot(agg='salary', data=emp, groupby='dept', kind='kde', orient='v')
<matplotlib.axes._subplots.AxesSubplot at 0x118cb77f0>
Splitting into separate plots
The row
and col
parameters can be used to split the data into separate plots. Each unique value of row
or col
will create a new plot. A one-item tuple consisting of the entire Figure is returned.
dxp.aggplot(agg='salary', data=emp, groupby='experience_level', kind='kde', orient='v', row='dept')
(<Figure size 576x1152 with 6 Axes>,)
Use the wrap
parameter to make new rows/columns
Set the wrap
parameter to an integer to determine where a new row/column will be formed.
dxp.aggplot(agg='salary', data=emp, groupby='experience_level', kind='box', orient='v', row='dept', wrap=3)
(<Figure size 864x720 with 6 Axes>,)
wrap
works for both row
or col
dxp.aggplot(agg='salary', data=emp, groupby='experience_level', kind='box', orient='v', col='dept', wrap=5)
(<Figure size 1296x576 with 6 Axes>,)
Use both row
and col
for a entire grid
By using both row
and col
, you can maximize the number of variables you divide the data into.
dxp.aggplot(agg='salary', data=emp, groupby='gender', kind='kde', row='dept', col='experience_level')
(<Figure size 1008x1152 with 18 Axes>,)
Normalize by more than one variable
Before, we normalized by just a single variable. It is possible to normalize by multiple variables with a tuple. Here we normalize by department and gender. Adding up all the blue bars for each department should add to 1.
dxp.aggplot(agg='dept', data=emp, hue='gender', kind='bar', row='race', normalize=('agg', 'hue'))
(<Figure size 720x1008 with 5 Axes>,)
Normalize by three variables
Here we normalize by race, experience level, and gender. Each set of orange/blue bars within each plot will add to 1.
dxp.aggplot(agg='dept', data=emp, hue='gender', kind='bar', row='race',
col='experience_level', normalize=('hue', 'col', 'row'), orient='h')
(<Figure size 1008x1008 with 15 Axes>,)
Scatterplot
scatterplot
is the only other currently available function. It plots two continuous valued variables against each other. It does not do any aggregating. It plots the data raw is it sees it. It can split the data into groups or new plots with hue
, row
, and col
.
dxp.scatterplot('experience', 'salary', data=emp)
<matplotlib.axes._subplots.AxesSubplot at 0x1271ec710>
Split data in the same plot with hue
dxp.scatterplot('experience', 'salary', data=emp, hue='gender')
<matplotlib.axes._subplots.AxesSubplot at 0x1274f1a20>
Plot a regression line by setting fit_reg
equal to True
By default it plots the 95% confidence interval around the mean.
dxp.scatterplot('experience', 'salary', data=emp, hue='gender', fit_reg=True)
<matplotlib.axes._subplots.AxesSubplot at 0x127670e10>
Further split the data into separate plots with row
and col
dxp.scatterplot('experience', 'salary', data=emp, hue='gender', row='dept', wrap=3)
(<Figure size 864x720 with 6 Axes>,)
dxp.scatterplot('experience', 'salary', data=emp, hue='gender', row='dept', col='experience_level')
(<Figure size 1008x1152 with 18 Axes>,)
Use the s
parameter to change the size of each marker
Let s
equal a column name containing numeric values to set each marker size individually. We need to create another numeric variable first since the dataset only contains two.
import numpy as np
emp['num'] = np.random.randint(10, 300, len(emp))
dxp.scatterplot('experience', 'salary', data=emp, hue='gender', row='dept', wrap=3, s='num')
(<Figure size 864x720 with 6 Axes>,)
Comparison with Seaborn
If you have used the Seaborn library, then you should notice a lot of similarities. Much of Dexplot was inspired by Seaborn. Below is a list of the extra features in Dexplot not found in Seaborn
- The ability to graph frequency percentage and normalize over any number of variables
- Far fewer public functions. Only two at the moment
- No need for multiple functions to do the same thing. Seaborn has both
countplot
andbarplot
- Ability to make grids with a single function instead of having to use a higher level function like
catplot
- Pandas
groupby
methods are available as strings - Both x/y-labels and titles are automatically wrapped so that they don't overlap
- The figure size (plus several other options) and available to change without dipping down into matplotlib
- No new types like FacetGrid. Only matplotlib objects are returned
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.