Jouni Helske
LightNews LightNews
banner
jounihelske.bsky.social
Jouni Helske
@jounihelske.bsky.social
640 followers 420 following 130 posts
Academy Research Fellow at INVEST (University of Turku), PI of CAUSALTIME project. Bayesian statistics, longitudinal causal inference, hidden Markov and state space models in general. Computational statistics in social sciences and various other things.
Posts Replies Media Videos
jounihelske.bsky.social
I'm often struggling with first sentences of Introduction when writing a paper. Perhaps I'm trying to be too fancy and should just go with "The paper is sociologically motivated. The content statistical." for one work in progress. #stats #metasci #sociology #AcademicSky
doi.org/10.1214/aoms...
Title and first sentence of the paper "Probabilistic functions of finite state Markov chains" by T. Petrie (1969): "This paper is statistically motivated; the content mathematical. The motivation is this: ..."
June 24, 2025 at 5:28 AM
jounihelske.bsky.social
I couldn't help myself, a quick simulation doesn't confirm my reasoning:

gist.github.com/helske/744d6...
April 29, 2025 at 5:53 PM
jounihelske.bsky.social
Congratulations to my first PhD student Tiia-Maria Pasanen who successfully defended her thesis "Bayesian spatio-temporal modeling of areal data" on Saturday. Excellent work with applications in economic history, epidemiology, and sociology. #stats #rstats #AcademicSky #econsky #sociology
January 13, 2025 at 11:09 AM
jounihelske.bsky.social
Just got the same review request for Journal of AppliedMath (that seems to be their official name) to my two different email addresses, send by seemingly different persons but signed by same, third person. Bit sketchy... #metasci #AcademicSky
Screenshot of two emails for requesting a review fore same article, with different senders.
December 7, 2024 at 9:21 AM
jounihelske.bsky.social
Nice Shiny app to create word clouds of your research based on the paper titles or abstracts (the version using Semantic Scholar). scholargoggler.com
A word cloud generated by https://scholargoggler.com/, showing words such as models, package, state, and space.
November 22, 2024 at 8:08 AM
jounihelske.bsky.social
Be careful in using as.integer(x) for numeric variable, use as.integer(round(x)) instead to avoid issues with floating point precision. #rstats #stats
R code and console output from set.seed(1) p <- runif(2) x <- c(1232, 439) y <- x / p z <- y * p x as.integer(z) as.integer(round(z)) showing how as.integer can lead to unexpected results.
November 11, 2024 at 1:06 PM
jounihelske.bsky.social
Computer scientists are getting a taste of their own medicine: after rebranding statistics as machine learning, AI is now physics. #NobelPrize #stats #metasci #AcademicSky🧪
Terminator with text "physics" standing in front of a desk while crying anime girl with text "machine learning" hides under the desk.
October 9, 2024 at 4:17 AM
jounihelske.bsky.social
Oh my. Rounding by statistical significance: Significant estimates are exactly correct, but those with large p-values are just random noise and are actually zero... #metasci #stats 📉📈
A line graph starting below zero, with some points exactly zero and ending above zero. Below figure there is a text "Note: Statistically insignificant effects are set to zero."
January 31, 2024 at 10:52 AM
jounihelske.bsky.social
Looks interesting, but I wouldn't put too much weight on this paper. 1) this is only about naming reviewers after reviews, not actually open comments. 2) There is likely still selection bias and confounding present, for example, authors might be more interested in naming bigger names as reviewers.
Screenshot of Nature's policy on open review: After acceptance, authors can opt to include the names of the reviewers with their consent.
January 10, 2024 at 6:36 AM
jounihelske.bsky.social
I managed to crash chatGPT for the first time, it just keeps generating na.rm = TRUE.
A screenshot of chatGPToutput containing R code which starts normally but then ends up repeating argument na.rm = TRUE infinitely.
November 29, 2023 at 6:34 PM
jounihelske.bsky.social
There is also some discussion about this in Andrew Gelman's blog: statmodeling.stat.columbia.edu/2011/10/25/h...
I don't find the actual post super informative, the comments by Radford Neal are nice, e.g.,
October 25, 2023 at 7:53 AM