2 min read

Top posts in 2020

2020/12/31

Attention Conservation Notice: these are the most popular posts, so if you cared you would already have seen them

According to Google Analytics, these were the most popular pages on this site in 2020:

  1. What have I got against the Shapiro-Wilk test? This 2019 post was popular because it has been linked on StackOverflow and by Hadley Wickham, which is why it’s at the top. I still think it’s worth reading.
  2. Weights in statistics “There are roughly three and half distinct uses of the term weights in statistical methodology…” I think this is an important post.
  3. You will probably not be eaten by a grue. I wrote an interpreter for a subset of BASIC to run the ‘Oregon Trail’ game. A toy solution to a toy problem.
  4. What is ‘Data Science Practice’? Advertising for a course design. It’s a good course, but the advertising is worthwhile only because the demarcation dispute is a real problem for statistics departments vs computer science.
  5. What’s the right proof of the Continuous Mapping Theorem? From 2015? I have no idea why this was popular. I stand by the content, but I don’t see why that many people should care.
  6. When the sky didn’t fall. Political propaganda, as part of a failed NZ referendum campaign. I’m not sure why I cared so much about the cannabis referendum. Maybe it’s that legalisation is the rare policy change that seems Pareto-improving – it doesn’t affect most people, and it helps a substantial minority. The lying liars who lied seriously misleading advertising on the ‘no’ campaign also got right up my nose.
  7. Multifactor interventions and interactions. The brilliant Susan Murphy has been fighting for factorial studies in addiction and mental illness. This is absolutely a good idea. However, I’ve seen (including in grant proposals) people misunderstanding the FAQs on interaction and factorial design from her research group. So I wrote this.
  8. MOAR survey regression models. I wrapped the VGAM package by Other Thomas (Thomas Yee) so it would handle survey designs; this makes many more regression models available for complex survey data.

The other two pages from the top ten were the root page for the site and the About page.