Anscombe's quartet is a classic example that illustrates why visualizing data is important. The quartet consists of four datasets with similar statistical properties. Each dataset has a series of x values and dependent y values. We will tabulate these metrics in an IPython notebook. However, if you plot the datasets, they look surprisingly different compared to each other.

Seaborn color palettes are similar to matplotlib colormaps. Color can help you discover patterns in data and is an important visualization component. Seaborn has a wide range of color palettes, which I will try to visualize in this recipe.

The matplotlib color maps are getting a lot of criticism lately because they can be misleading; however, most colormaps are just fine in my opinion. The defaults are getting a makeover in matplotlib 2.0 as announced at http://matplotlib.org/style_changes.html (retrieved July 2015). Of course, there are some good arguments that do not support using certain matplotlib colormaps, such as jet. In art, as in data analysis, almost nothing is absolutely true, so I leave it up to you to decide. In practical terms, I think it is important to consider how to deal with print publications and the various types of color blindness. In this recipe, I visualize relatively safe colormaps with colorbars. This is a tiny selection of the many colormaps in matplotlib.

Interactive IPython notebook widgets are, at the time of writing (July 2015), an experimental feature. I, and as far as I know, many other people, hope that this feature will remain. In a nutshell, the widgets let you select values as you would with HTML forms. This includes sliders, drop-down boxes, and check boxes. As you can read, these widgets are very convenient for visualizing the weather data I introduced in Chapter 1, Laying the Foundation for Reproducible Data Analysis.

If you don't have many variables in your dataset, it is a good idea to view all the possible scatterplots for your data. You can do this with one function call from either seaborn or pandas. These functions display a matrix of plots with kernel density estimation plots or histograms on the diagonal.

D3.js is a JavaScript data visualization library released in 2011, which we can also use in an IPython notebook. We will add hovering tooltips to a regular matplotlib plot. As a bridge, we need the mpld3 package. This recipe doesn't require any JavaScript coding whatsoever.

$ [sudo] pip install mpld3

Heat maps visualize data in a matrix using a set of colors. Originally, heat maps were used to represent prices of financial assets, such as stocks. Bokeh is a Python package that can display heatmaps in an IPython notebook or produce a standalone HTML file.

Violin plots combine box plots and kernel density plots or histograms in one type of plot. Seaborn and matplotlib both offer violin plots. We will use Seaborn in this recipe on z-scores of weather data. The z-scoring is not essential, but without it, the violins will be more spread out.

A hive plot is a visualization technique for plotting network graphs. In hive plots, we draw edges as curved lines. We group nodes by some property and display them on radial axes. NetworkX is one of the most famous Python network graph libraries; however, it doesn't support hive plots yet (July 2015). Luckily, several libraries exist that specialize in hive plots. Also, we will use an API to partition the graph of Facebook users available at https://snap.stanford.edu/data/egonets-Facebook.html (retrieved July 2015). The data belongs to the Stanford Network Analysis Project (SNAP), which also has a Python API. Unfortunately, the SNAP API doesn't support Python 3 yet.