Hit and Miss fifth anniversary data analysis

Because aside from writing a weekly newsletter, my hobbies include analysing data, especially my own

For the fifth anniversary of Hit and Miss, I took the archive as a corpus and did some analysis. (The corpus excludes the fifth anniversary issue, just to be clear!)

A key part of the newsletter is the links. (You can check out the analysis code for this section. No fancy RMarkdown, though, I’m not there yet!) There were many: 1,546 links in 261 issues. These include 1,416 distinct links:

  • tinyletter.com/UniversityOfWinds, digital.canada.ca, and lucascherkewski.com/study/franklin-definitions/ were the most common, each appearing about 6 times. This makes sense! I really admire Mita Williams’s newsletter—and recommend you subscribe, but it has no public archive, so I only ever link to the subscription page; CDS is where I work, and I need to wave at that sometimes; Ursula Franklin’s set of pacifist definitions continue to push me to pause and consider.
  • lucascherkewski.com was the most common domain, unsurprisingly, appearing 161 times. Twitter (157), CBC (58), The Globe and Mail (50) all follow (I have a hunch that links to the Globe started going up after I subscribed in the fall of 2020).
  • aworkinglibrary.com was the 6th most common domain (tied with GitHub and YouTube)—but notable for being the most common personal / non-institutional domain. No surprise, as Mandy’s writing is a huge inspiration for what I try to do with my newsletter.

Almost every issue has links. Only 5% of issues—13 of 261—had no links:

  • Interestingly, 10 of those 13 link-less issues were in the first 100 issues (first 2 years) of the newsletter, and it’s now been over a year since I sent an issue without links (though, amusingly, it noted that “I’ve accumulated a fair number of links to share, but they’ll have to wait for next week”). This suggests that earlier issues were more likely to deal exclusively with my offline reading and thinking, while later issues more consistently include a link.
  • If we take out newsletters that only had “selflinks” (i.e., the only links in the newsletters were to a page on my website, like another issue of the newsletter), this number grows to 8% of issues, or 22 of 261. Issue 219, “Silent commemoration” is a good example of selflinking, with links only to pages on my site and a previous issue.

The number of links varies pretty significantly from issue to issue!


The other key part of the newsletter is its words! (You can check out the word analysis code!) In the interest of time, I haven’t done a full analysis, but some interesting observations:

There’s more I could do for a proper text analysis (TF-IDF, clustering, topic models, etc), but we’ll leave that “for future study”. :)