Yankee Candle reviews on Amazon

Suggested answers

Part 1 - Data scraping

See yankee-candle-scrape-A.R for suggested scraping code.

Part 2 - Data analysis

Let’s start by loading the packages we will need:

library(tidyverse)
library(lubridate)

• Load the data you saved into the data folder and name it yc_reviews_all.

yc_reviews_all <- read_csv("data/yc-reviews-all.csv")

• Your turn (5 minutes): Our goal is to plot the number of reviews or titles mentioning “no scent” or “no smell” per week. Take a look at the data you loaded and outline how you can transform it to prepare it for such a plot. Your response should be bullet points that can roughly be translated to steps in a data wrangling pipeline.
  • Extract date of review.
  • Figure out which week of the year the date corresponds to.
  • Look for “no scent” or “no smell” in the titles or reviews.
  • Count how many titles or reviews have these phrases mentioned for each week.
  • Plot the number of mentions by week against week number.
• Demo: Create the plot described above.

yc_reviews_all |>
  mutate(
    date = mdy(country_date),
    week = week(date),
    review = str_to_lower(review),
    title = str_to_lower(title),
    no_scent = case_when(
      str_detect(review, "no scent") | str_detect(title, "no scent") ~ "mentioned",
      str_detect(review, "no smell") | str_detect(title, "no smell") ~ "mentioned",
      TRUE ~ "not mentioned"
    )
  ) |>
  count(week, no_scent) |>
  filter(no_scent == "mentioned") |>
  ggplot(aes(x = week, y = n, group = 1)) +
  geom_line()

• Your turn (3 minutes): Outline how this exercise could be expanded to make a plot like this one that compares number of reviews mentioning no scent/smell and number of COVID cases.

Scrape data going further back and join it with weekly COVID cases data from another source and plot both.