What is machine learning, and how does it work?

Broadly speaking, machine learning uses computer programs to identify patterns in data. In many ways, these techniques automate tasks that researchers have done by hand for years. For instance, we often ask people taking our surveys to tell us, in their own words, how they feel about a particular issue. A researcher then reads those responses, identifies key themes and organizes those themes for analysis. But what if we wanted to analyze several million social media posts? Or tens of thousands of lengthy congressional speeches? That level of scale and complexity is much more difficult for humans to process. In these types of situations, machine learning can be very useful. In the latest installment of our Methods 101 video series, our Data Labs team explains what machine learning is and how our researchers use it.

How Pew Research Center is covering COVID-19

As an organization committed to informing the public by providing facts and analysis about the most pressing issues of the day, Pew Research Center pivoted in the early days of the U.S. coronavirus outbreak to cover the far-reaching impact that the global pandemic is having on our society. We have enlisted staff from all research areas at the Center to capture multiple facets of how the coronavirus pandemic is impacting life at the local, national and international level – and since mid-March, we have produced more than 80 reports and articles. The Center’s response to the COVID-19 crisis has also included the difficult decision to suspend much of our international survey work until further notice because most of our international surveys are still conducted face-to-face. (Our polling in countries where surveys are conducted by phone is continuing, as is our U.S. polling, which is conducted primarily online.)

Introducing pewmethods: An R package for working with survey data

In March, Pew Research Center’s survey methods team proudly released version 1.0 of pewmethods, an R package containing various functions that we use in our day-to-day survey data work. The package was originally envisioned as an internal way to reuse, maintain and share code – but since many of the problems that these functions were designed to solve are not unique to our projects, we made the pewmethods package publicly available on Github for other researchers. In the blog posts listed below, learn how the pewmethods package can help with creating basic survey weights and carrying out various cleaning, recoding, combining and collapsing tasks. You can also use the package to analyze our international survey data:

• Exploring survey data with the pewmethods R package
• Weighting survey data with the pewmethods R package
• Analyzing international survey data with the pewmethods R package

Pew Research Center at AAPOR 2020

On Thursday, June 11 and Friday, June 12, Pew Research Center experts will present at the annual conference of the American Association of Public Opinion Research, held virtually. Conference attendees interested in seeing Center researchers speak about developments in research methodology and recent findings can find a schedule of presentations here. Follow @pewmethods and #AAPOR on Twitter to keep up with the event in real time, and watch the all-attendee kickoff session videos to hear Vice President of Research Claudia Deane talk about our COVID-19 survey work.

Have you seen our American News Pathways data tool?

Our yearlong American News Pathways project focuses how Americans' news habits and attitudes relate to what they hear, perceive and know about the 2020 U.S. election campaign and the COVID-19 outbreak. Use our interactive data tool to examine election- and coronavirus-related survey questions by media trait or demographic group. You can also download our datasets connect to our API.

New on 'Decoded'

Pew Research Center’s Decoded blog focuses on the “how” behind our numbers. The blog features content ranging from survey methods, to data science, to data visualization, and allows researchers to build on and engage with our work. Explore some of our latest posts:

• Communicating technical content
• The challenges of using real-time epidemiological data in a public health crisis
• The ‘class size paradox’: How individual- and group-level perspectives differ, and why it matters in research

Featured datasets