11.6. Creating Plots Using plotly

In this section, we cover the basics of the plotly Python package, the main tool we use in this book to create plots.

The plotly package has several advantages over other plotting libraries. It creates interactive plots rather than static images. When you create a plot in plotly, you can pan and zoom to see parts of the plot that are too small to see normally. You can also hover over plot elements, like the symbols in a scatter plot, to see the raw data values. Also, plotly can save plots using the SVG file format, which means that images appear sharp even when zoomed in. If you’re reading this chapter in a PDF or paper copy of the book, we used this feature to render plot images. Finally, it has a simple “express” API for creating basic plots, which helps when you’re doing exploratory analysis and want to quickly create many plots.

We go over the fundamentals of plotly in this section. We recommend using the official plotly documentation if you encounter something that isn’t covered here 1.

11.6.1. Figure and Trace Objects

Every plot in plotly is wrapped in a Figure object. Figure objects keep track of what plots to draw. For instance, a single Figure can draw a scatter plot on the left and a line plot on the right. Figure objects also keep track of the plot layout, which includes the size of the plot, title, legend, and annotations.

Let’s look at an example using the dataset of dog breeds.

The plotly.express module provides a concise API for making plots.

import plotly.express as px

We use plotly.express below to make a scatter plot of weight against height for the dog breeds.
Notice that the return value from .scatter() is a Figure object.

fig = px.scatter(dogs, x='height', y='weight', 
                 width=350, height=250)

# fig is a plotly Figure object:
fig.__class__
plotly.graph_objs._figure.Figure

Displaying a Figure object renders it to the screen.

fig
../../_images/viz_plotly_12_0.svg

This particular Figure holds one plot, but Figure objects can hold any number of plots. Below, we create a facet of three scatter plots.

fig = px.scatter(dogs, x='height', y='weight',
                 facet_col='size',
                 width=650, height=250)
fig.update_layout(margin=dict(t=30))
fig
../../_images/viz_plotly_14_0.svg

These three plots are stored in Trace objects. However, we don’t usually manipulate Trace objects manually. Instead, plotly provides functions that automatically create facetted subplots, like the px.scatter function we used here. Now that we have seen how to make a simple plot, we next show how to modify plots.

11.6.2. Modifying Layout

We often need to change a figure’s layout. For instance, we might want to adjust the figure margins or change the axis range. To do this, we can use the Figure.update_layout() method. Let’s look at an example of a scatter plot where the title is cut off because the plot doesn’t have large enough margins.

fig = px.scatter(dogs, x='weight', y='longevity',
                 title='Smaller dogs live longer',
                 width=350, height=250)
fig
../../_images/viz_plotly_18_0.svg

We can adjust the margin to give enough space for the title.

fig = px.scatter(dogs, x='weight', y='longevity',
                 title='Smaller dogs live longer',
                 width=350, height=250)

fig.update_layout(margin=dict(t=30))
fig
../../_images/viz_plotly_20_0.svg

The .update_layout() lets us modify any property of a layout. This includes the plot title (title), margins (margins dictionary), and whether to display a legend (showlegend). The plotly documentation has the full list of layout properties 2.

Figure objects also have .update_xaxes() and .update_yaxes() functions, which are similar to .update_layout(). These two functions let us modify properties of the axes, like the axis limits (range), number of ticks (nticks), and axis label (title). Below, we adjust the range of the y-axis and change the title on the x-axis.

fig = px.scatter(dogs, x='weight', y='longevity',
                 width=350, height=250)

fig.update_yaxes(range=[5, 18])
fig.update_xaxes(title='average weight (lb)')
fig
../../_images/viz_plotly_22_0.svg

The plotly package comes with many plotting methods; we describe several of them in the next section.

11.6.3. Plotting Functions

The plotly methods includes line plots, scatter plots, bar plots, box plots, and histograms. The API is similar for each type of plot. The data frame is the first argument. Then, we can specify a column of the data frame to place on the x-axis and a column to place on the y-axis using the x and y keyword arguments.

We begin with a line plot of median time each year for the runners in the Cherry Blossom race.

px.line(medians, x='year', y='time',
        width=350, height=250)
../../_images/viz_plotly_27_0.svg

Next, we make a bar plot of average longevity for different size dog breeds.

lifespans = dogs.groupby('size')['longevity'].mean().reset_index()

px.bar(lifespans, x='size', y='longevity',
       width=350, height=250)
../../_images/viz_plotly_29_0.svg

Plotting methods in plotly also contain arguments for making facet plots. We can facet using color on the same plot, plotting symbol or line style.
Or we can facet into multiple subplots. Below are examples of each. We first make a scatter plot of height and weight of dog breeds and use different plotting symbols and colors to facet within the plot by size.

fig = px.scatter(dogs, x='height', y='weight', 
                 color='size', symbol='size',
                 width=350, height=250)
fig
../../_images/viz_plotly_31_0.svg

The plot below shows side-by-side histograms of longevity for each breed size. Here we facet by columns.

fig = px.histogram(dogs, x='longevity', facet_col='size',
                   width=550, height=250)
fig.update_layout(margin=dict(t=30))
../../_images/viz_plotly_33_0.svg

For the complete list of plotting functions, see the main documentation for plotly 1 or plotly.express 3, the submodule of plotly that we primarily use in the book.

To add context to a plot, we use the plotly annotation methods; these are described next.

11.6.4. Annotations

The Figure.add_annotation() method places annotations on a plotly figure. These annotations are line segments with text and an optional arrow. The location of the arrow is set using the x and y parameters, and we can shift the text from its default position using the ax and ay parameters. Below, we annotate the scatter diagram with information about one of the points.

fig = px.scatter(dogs, x='weight', y='longevity',
                 width=350, height=250)

fig.add_annotation(text='Chihuahuas live 16.5 years on average!',
                   x=2, y=16.5,
                   ax=30, ay=5,
                   xshift=3,
                   xanchor='left')
fig
../../_images/viz_plotly_36_0.svg

This section covered the basics of creating plots using the plotly Python package. We introduced the Figure object, which is the object plotly uses to store plots and their layouts. We covered the basic plot types that plotly makes available, and a few ways to customize plots by adjusting the layout and axes, and by adding annotations. In the next section, we briefly compare plotly to other common tools for creating visualizations in Python.


1(1,2)

https://plotly.com/python/

2

https://plotly.com/python-api-reference/generated/plotly.graph_objects.Layout.html

3

https://plotly.com/python-api-reference/plotly.express.html