June 4, 2025
Briefly
- Ask Stuff’s BudgetBot anything you need to know about Budget 2025. Or, perhaps, don’t. The budget is an occasion where there’s actually quite a lot of expert analysis both in mainstream media and social media, as well as all the unreliable vibes-based commentary anyone could want
- I’m not saying that AI is useless. Mike Caulfield has some very impressive demonstrations of forcing Claude to help with fact checking.
- Various sources report that smartphones cause haemorrhoids. This is based on unpublished research presented at a conference, of a sort that can’t show more than a correlation, and that just barely provides evidence of a correlation (this is the closest we have to details). The problem is that the study has no ability to tell whether using a smartphone on the toilet increases the risk of hemorrhoids, or having hemorrhoids increases the chance that you’ll use a smartphone on the toilet, or that something else affects both things. Also, it’s about time for this story to reappear, since we had it in 2023 and 2020 and 2018, though those times it was at least presented as expert opinion rather than research.
- RadioNZ have launched a political poll with Reid Research (hat tip to @danxduran on Bluesky). Notably, the article describes not just the maximum margin of error, for proportions near 50%, but margins of error for smaller proportions such as 10% or 20% (not for 5%, unfortunately). These are uncertainty estimates for an idealised mathematical model of polling, and underestimate the true uncertainty a bit, but they are a big step forward. I’ve written about the uncertainty for smaller probabilities on StatsChat before
Thomas Lumley (@tslumley) is Professor of Biostatistics at the University of Auckland. His research interests include semiparametric models, survey sampling, statistical computing, foundations of statistics, and whatever methodological problems his medical collaborators come up with. He also blogs at Biased and Inefficient See all posts by Thomas Lumley »