Null Hypothesis #1

Hypothesis

“Americans and Russians have different viewing habits on YouTube”.

Dataset

Trending YouTube Video Statistics, including

• 40949 trending videos from US published between 2006–07–23 and 2018–06–14, and
• 40739 trending videos from RU published between 2011–06–20 and 2018–06–14.

Source: Kaggle, https://www.kaggle.com/datasnaek/youtube-new

Analysis

A little cross-tabulation.

  table(videos$category, videos$nation)
                          RU    US
  
  Autos & Vehicles       1583   384
  Comedy                 3065  3457
  Education               714  1656
  Entertainment          5943  9964
  Film & Animation       3041  2345
  Gaming                 1043   817
  Howto & Style          2000  4146
  Movies                    1     0
  Music                  1895  6472
  News & Politics        5402  2487
  Nonprofits & Activism     0    57
  People & Blogs        10350  3210
  Pets & Animals          604   920
  Science & Technology   1133  2401
  Shows                   194    57
  Sports                 1968  2174
  Travel & Events         262   402

Removing two categories with zeroes (“Movies” and “Nonprofits & Activism”) to avoid errors with Chisq Test.

Get Giancarlo Vercellino’s stories in your inbox

Join Medium for free to get updates from this writer.

Remember me for faster sign in

Selecting the same time frame both for RU and US trending videos (that is the window from 2011–06–20 to 2018–06–14).

time_index <- as.Date(videos$publish_time) < as.Date(tail(RUvideos$publish_time, 1)) & as.Date(videos$publish_time) > as.Date(head(RUvideos$publish_time, 1))
excluded <- c("Movies", "Nonprofits & Activism")
category_index <- !(videos$category %in% excluded)
table(videos$category[category_index & time_index], videos$nation[category_index & time_index])
                           RU    US
  Autos & Vehicles       1557   347
  Comedy                 3031  3402
  Education               704  1593
  Entertainment          5866  9743
  Film & Animation       3007  2261
  Gaming                 1022   806
  Howto & Style          1979  4086
  Music                  1876  6319
  News & Politics        5345  2414
  People & Blogs        10182  3132
  Pets & Animals          596   904
  Science & Technology   1112  2321
  Shows                   191    57
  Sports                 1948  2107
  Travel & Events         253   391

Let’s do the test.

chisq.test(videos$nation[category_index & time_index], videos$category[category_index & time_index])
Pearson's Chi-squared test
data:  videos$nation[category_index & time_index] and videos$category[category_index & time_index]
X-squared = 10791, df = 14, p-value < 2.2e-16

Comment

American’s and Russian’s trending videos on YouTube show possibly different viewing habits (Pearson’s Chi-squared test: p-value < 2.2e-16; the p-value significance does not fade even removing extreme quantiles of video views): it seems that Americans have a prevalent interest on Education, Entertainment, Howto & Style, Music, Pets & Animals, Travel & Events, while Russians have a prevalent interest on Autos & Vehicles, Film & Animation, Gaming, News & Politics, People & Blogs, Shows. In many cases, the differences between the two groups are remarkable.

Next

In the next issue of Null Hypothesis, we will go deep into the rabbit hole comparing the different viewing habits between Americans and Russians with reference to News & Politics. Scary!