1 Project Approach

For this assignment, I plan to use the New York Times Most Popular API to explore which news sections appear most frequently among the most-viewed articles over the past 30 days. My main research question is: Which New York Times news sections are most represented among popular articles? I want to identify the top five sections and compare how frequently they appear in the results.

I will use R to connect to the API using a securely stored API key and retrieve the data in JSON format. I will then convert the JSON response into a tidy data frame containing article titles, sections, publication dates, and authors. I plan to use packages such as httr2, dplyr, and ggplot2 to collect, clean, analyze, and visualize the data.

Some challenges I expect include missing information, nested JSON fields, API request limits, and differences in how news sections are labeled. I will handle missing values, check for duplicate articles, and make sure the data is organized correctly before analyzing it. Finally, I plan to create a simple bar chart comparing the most frequently represented sections and document my findings in an R Markdown report.

2 Introduction

The New York Times provides several APIs for accessing article information. For this assignment, I selected the Most Popular API, specifically the Most Viewed endpoint, which returns metadata about articles that received high readership during a selected period.

My objective is to identify which news sections are most frequently represented among the returned articles and compare their distribution.

API Endpoint:

https://api.nytimes.com/svc/mostpopular/v2/viewed/30.json

The 30 in the endpoint specifies the 30-day period. The request also uses an api-key query parameter for authentication.

3 Setup

The following packages are used to connect to the API, process JSON data, organize the results, and create visualizations.

library(httr2)
library(dplyr)
library(purrr)
library(tibble)
library(ggplot2)
library(knitr)

4 API Authentication and Data Collection

To protect my API key, I stored it in the .Renviron file and accessed it using Sys.getenv(). This avoids including sensitive credentials directly in the report.

# Retrieve API key from environment
api_key <- Sys.getenv("NYT_API_KEY")

if (!nzchar(api_key)) {
  stop("NYT_API_KEY is missing. Configure it in .Renviron.")
}

# Define the API endpoint
endpoint <- "https://api.nytimes.com/svc/mostpopular/v2/viewed/30.json"

# Send request to NYT API
response <- request(endpoint) |>
  req_url_query(`api-key` = api_key) |>
  req_perform()

# Confirm successful request
status_code <- resp_status(response)

cat("API Response Status:", status_code)
## API Response Status: 200

5 Parsing the JSON Response

The API returns data in JSON format. I used resp_body_json() to parse the response into an R object.

# Convert JSON response to R object
nyt_data <- resp_body_json(
  response,
  simplifyVector = FALSE
)

# Check the API response
cat("API Status:", nyt_data$status, "\n")
## API Status: OK
cat("Articles Returned:", length(nyt_data$results))
## Articles Returned: 20

6 Creating a Tidy Data Frame

The JSON response contains nested information about each article. I extracted six fields: article ID, title, section, publication date, author, and URL.

# Extract article fields into a tibble
articles <- map_dfr(nyt_data$results, function(x) {
  tibble(
    article_id = as.character(x$id),
    title = x$title,
    section = x$section,
    published_date = as.Date(x$published_date),
    author = x$byline,
    url = x$url
  )
})

# Display dimensions
cat("Rows:", nrow(articles), "\n")
## Rows: 20
cat("Columns:", ncol(articles), "\n")
## Columns: 6
# Display first 10 records
kable(
  head(articles, 10),
  caption = "Sample of NYT Most-Viewed Articles"
)
Sample of NYT Most-Viewed Articles
article_id title section published_date author url
100000011143994 Latest Polls Polls 2026-09-10 https://www.nytimes.com/interactive/polls/latest-polls.html
100000007018475 Patrick Stewart, Master of Shakespeare, ‘Star Trek’ and More, Has Died Arts 2026-10-29 By Anita Gates and William McDonald https://www.nytimes.com/2026/10/29/arts/television/patrick-stewart-dead.html
100000011159452 She Was Offered $40 Million for Silence on Elon Musk. She Spoke Out Instead. Movies 2026-10-09 By Coralie Kraft https://www.nytimes.com/2026/10/09/movies/ashley-st-clair-interview-elon-musk.html
100000011238489 Obama’s ‘Cheesecake Factory’ Jab Rankles Trump on Mississippi Campaign Stop U.S. 2026-10-26 By Chris Cameron https://www.nytimes.com/2026/10/26/us/politics/obamas-cheesecake-factory-jab-rankles-trump-on-mississippi-campaign-stop.html
100000011066985 It Can Keep Cancer in Remission. Why Is No One Talking About It? Well 2026-10-06 By Nina Agrawal https://www.nytimes.com/2026/10/06/well/cancer-prevention-exercise-treatment.html
100000011191476 How Cornell Punished Each of the 7 Men Accused of Sexual Assault New York 2026-10-03 By Sharon Otterman, Hurubie Meko and Debra Kamin https://www.nytimes.com/2026/10/03/nyregion/cornell-men-university-discipline.html
100000010940076 O.J. Simpson, a Secret Informant and a Decades-Old Mystery U.S. 2026-10-18 By Mike Baker and Tim Arango https://www.nytimes.com/2026/10/18/us/oj-simpson-ippolito-informant-accomplice.html
100000011213021 Trump’s New Press Secretary: Not-Quite Attorney at Law U.S. 2026-10-14 By Shawn McCreesh https://www.nytimes.com/2026/10/14/us/politics/trump-press-secretary-katie-zacharia-lawyer.html
100000011227959 Obama Issues Searing Attacks on Trump and Ramaswamy in Return to Campaign Trail U.S. 2026-10-24 By Tim Balk, Emily Davies, John Yoon and Richard Fausset https://www.nytimes.com/2026/10/24/us/politics/obama-trump-midterms-campaign.html
100000011127139 They Argued About Trump at a Dog Park. Then They Pulled Their Guns. U.S. 2026-10-05 By Jessica Contrera https://www.nytimes.com/2026/10/05/us/trump-ohio-dog-park-shooting.html

7 Data Cleaning and Validation

Before analyzing the results, I checked for missing values, blank section names, and duplicate article IDs. I also standardized section names by removing unnecessary whitespace.

# Check missing values
missing_values <- colSums(is.na(articles))

kable(
  data.frame(
    Column = names(missing_values),
    Missing_Values = as.integer(missing_values)
  ),
  caption = "Missing Values by Column"
)
Missing Values by Column
Column Missing_Values
article_id 0
title 0
section 0
published_date 0
author 0
url 0 
# Check duplicates before cleaning
duplicate_count <- sum(duplicated(articles$article_id))

cat("Duplicate Article IDs:", duplicate_count, "\n")
## Duplicate Article IDs: 0
# Clean data
articles_clean <- articles |>
  distinct(article_id, .keep_all = TRUE) |>
  mutate(
    section = trimws(section),
    section = if_else(
      is.na(section) | section == "",
      "Unknown",
      section
    ),
    author = na_if(trimws(author), "")
  )

cat("Cleaned Articles:", nrow(articles_clean))
## Cleaned Articles: 20

7.1 Data-Cleaning Decisions

I retained the fields most relevant to the research question and excluded nested multimedia information because it was unnecessary for the analysis. I converted publication dates into R’s Date format, checked for missing values, replaced missing or blank section names with “Unknown,” and removed duplicate article IDs. Blank author names were converted to missing values.

9 Data Visualization

The following chart compares the number of most-viewed articles in each section.

ggplot(
  section_summary,
  aes(
    x = reorder(section, article_count),
    y = article_count
  )
) +
  geom_col(fill = "steelblue") +
  coord_flip() +
  labs(
    title = "Most-Viewed NYT Articles by Section",
    subtitle = "Past 30 Days",
    x = "News Section",
    y = "Number of Articles"
  ) +
  theme_minimal()

10 Findings

The API returned 20 articles across 8 news sections.

The section with the highest representation was U.S., containing 11 articles, or 55% of the returned articles.

The second most represented section was Opinion, with 3 articles.

The five most represented sections in the current results were: U.S., Opinion, Arts, Movies, New York.

Based on the API results collected during testing, U.S. articles represented 55% of the 20 returned articles, followed by Opinion at 15%. The other six sections each represented 5%. These figures may change when the report is run again.

11 Conclusion

The analysis demonstrates how an API can be used to retrieve semi-structured JSON data and transform it into a clean data frame for analysis. The results indicate that the U.S. section had the highest representation among the most-viewed articles returned during testing.

However, the API only returns a selected list of popular articles, not all published articles or their actual page-view counts. Therefore, the findings describe section representation within the returned list rather than overall readership.

In the future, I could collect results at regular intervals and compare changes in the most frequently represented sections over time.

12 References

New York Times. (2026). Most Popular API. https://developer.nytimes.com/docs/most-popular-product/1/overview