Overview

This chapter presents a practical introduction to the main libraries that most statistics practitioners use in Python. It will cover some of the most important and useful concepts, functions, and Application Programming Interfaces (APIs) of each of the key libraries. Almost all of the computational tools that will be needed for the rest of this book will be introduced in this chapter.

By the end of this chapter, you will understand the idea behind array vectorization of the NumPy library and be able to use its sampling functionalities. You'll be able to initialize pandas DataFrames to represent tabular data and manipulate their content. You'll also understand the importance of data visualization in data analysis and be able to utilize Python's two most popular visualization libraries: Matplotlib and Seaborn.

After going through a refresher on the Python language in the previous chapter, we are now ready to tackle the main topics of this book: mathematics and statistics.

Among others, the general fields of computational mathematics and statistics can be broken up into three main tool-centric components: representation and engineering; analysis and computation; and finally, visualization. In the ecosystem of the Python programming language, specific libraries are dedicated to each of these components (namely, pandas, NumPy, Matplotlib, and Seaborn), making the process modular.

While there might be other similar packages and tools, the libraries that we will be discussing have been proven to possess a wide range of functionalities and support powerful options in terms of computation, data processing, and visualization, making them some of a Python programmer's preferred tools over the years.

In this chapter, we will be introduced to each of these libraries and learn...

The term scientific computing has been used several times in this workshop so far; in the broadest sense of the term, it denotes the process of using computer programs (or anything with computing capabilities) to model and solve a specific problem in mathematics, engineering, or science. Examples may include mathematical models to look for and analyze patterns and trends in biological and social data, or machine learning models to make future predictions using economic data. As you may have already noticed, this definition has a significant overlap with the general fields of data science, and sometimes the terms are even used interchangeably.

The main workhorse of many (if not most) scientific computing projects in Python is the NumPy library. Since NumPy is an external library that does not come preinstalled with Python, we need to download and install it. As you may already know, installing external libraries and packages in Python can be...

If NumPy is used on matrix data and linear algebraic operations, pandas is designed to work with data in the form of tables. Just like NumPy, pandas can be installed in your Python environment using the pip package manager:

$ pip install pandas

If you are using Anaconda, you can download it using the following command:

$ conda install pandas

Once the installation process completes, fire off a Python interpreter and try importing the library:

>>> import pandas as pd

If this command runs without any error message, then you have successfully installed pandas. With that, let's move on with our discussions, beginning with the most commonly used data structure in pandas, DataFrame, which can represent table data: two-dimensional data with row and column labels. This is to be contrasted with NumPy arrays, which can take on any dimension but do not support labeling.

Initializing a DataFrame Object

There are multiple...

Data visualization is undoubtedly an integral part of any data pipeline. Good visualizations can not only help scientists and researchers find unique insights about their data, but also help convey complex, advanced ideas in an intuitive, easy to understand way. In Python, the backend of most of the data visualization tools is connected to the Matplotlib library, which offers an incredibly wide range of options and functionalities, as we will see in this upcoming discussion.

First, to install Matplotlib, simply run either of the following commands, depending on which one is your Python package manager:

$ pip install matplotlib
$ conda install matplotlib

The convention in Python is to import the pyplot package from the Matplotlib library, like so:

>>> import matplotlib.pyplot as plt

This pyplot package, whose alias is now plt, is the main workhorse for any visualization functionality in Python and will therefore...

This chapter went through the core tools for data science and statistical computing in Python, namely, NumPy for linear algebra and computation, pandas for tabular data processing, and Matplotlib and Seaborn for visualization. These tools will be used extensively in later chapters of this book, and they will prove useful in your future projects. In the next chapter, we will go into the specifics of a number of statistical concepts that we will be using throughout this book and learn how to implement them in Python.