Did you spend 4-6 weeks writing your last analysis plan then get mocked by statisticians on the internet for being slow 😜. We had a go doing it with AI - still takes 4-6 weeks but the AI will do the boring parts letting you conentrate on the stats. Pre print-here bsky.app/profile/gfor...

We’ve just released a pre-print reporting the development and validation of an AI tool to write statistical analysis plans from protocols. www.linkedin.com/feed/update/... Thread 👇

fucking FINALLY

We've got something cooking along thse lines. Anyone else?

Having complicated feelings about this one…

Hey #rstats, What's your rule for splitting R scripts that form part of a wider analysis pipeline / project? I usually write a single script which includes sections for each step from data cleaning to the final results, but it can become unwieldy when the script becomes long. ...

New paper posted on Arxiv: "When do composite estimands answer non-causal questions?" This can happen more often than you think, and can have a dramatic impact on trial results (e.g. a false-positive rate of almost 90%) arxiv.org/abs/2506.22610 @timpmorris.bsky.social

On tabular health data, time and time again, I see linear (or generalised linear models) perform as well or better than machine learning algorithms that avoid linearity assumptions. I am surprised by this, as the linearity assumption is unlikely to be true. Does anyone else see this? Why is this?

Simpson's paradox

Georgi Baklicharov asks: can treatment effect testing in trials with intercurrent events be nearly assumption-free? #EuroCIM2025

Happy to see that ordered beta regression reached 100 citations on Google Scholar! The model has citations from work in climate science, ecology, medicine, psychology, & political science, just to name a few. Thanks to all of you for using ordbetareg (or glmmTMB)!! #rstats

I'm working with data that I'm not allowed to share -- I'd like to generate synthetic data so that others at least have something that they can run my code on! Any pointers to tutorials, favorite packages etc.?

Check out my new screencast, where I walk through how I use #Positron for #rstats package development work. I decided to release a new version of an R package to CRAN ✨live✨ this time around! youtu.be/uL3NZQIMrpk

In coursework, the contrast between ridge regression and the LASSO is really emphasized. After all, the latter actually does feature selection, by virtue of having a sparse solution to the minimum l1 problem, versus ridge's nonsparse solution in l2, pretty cool. 1/2

R packages for consort diagrams

We tried to look at ways to obtain flexible calibration plots in clustered (e.g. multicenter) validation studies. Work with @lasaibarrenada.bsky.social @laurewynants.bsky.social @bavodccampo.bsky.social arxiv.org/abs/2503.08389

The term "digital twin" as it is now used in medicine has no real relationship to how the term is used in engineering. Yet every paper on the former talks about the success of the latter as if that's relevant. They are not the same!

My spell check is trying to change bootstrap to Boomer. As in Boomer p-values. I didn't know the generation wars had made it to statistical inference. What next? Millennial credible intervals?

I think you mean "non-specific".

It's amazing how far back ideas go. This paper from 1984 by @f2harrell.bsky.social discusses the need for train/test data splits and the importance of assessing model calibration and discrimination and instability in variable selection methods. onlinelibrary.wiley.com/doi/epdf/10....

A minor stylistic preference I’ve recently found myself using: When introducing a key initialism or acronym in a paper, put the compressed version in the text and its expansion in parentheses. Instead of ‘under missing at random (MAR)’, use ‘under MAR (missing at random)’. 1/

What advice do folks have for organising projects that will be deployed to production? How do you organise your directories? What do you do if you're deploying multiple "things" (e.g. an app and an api) from the same project?

Just published a couple of pre-prints for those interested in sample size calculations for precise and fair individual-level predictions ... (not the end of the story, but a useful contribution we hope): Binary outcomes: arxiv.org/abs/2407.09293 Survival outcomes: arxiv.org/abs/2501.14482

Am looking for a particular article that argued that results of "explanatory" trials might generally be more timeless than those of pragmatic trials, since the latter might be reliant on a particular context at a given moment in time. Can't remember who this was by - something from imp science lit?

'Estimating hypothetical estimands with causal inference and missing data estimators in a diabetes trial case study', led by Camila Olarte Parra, with Rhian Daniel and David Wright, now available open access in Biometrics: doi.org/10.1093/biom...

#statstab #267 That’s a Lot to Process! Pitfalls of Popular Path Models Thoughts: The number of senior psych researchers that just loooove Process...and all use it post hoc 😩 Causal inference is not easy. #mediation #process #spss #causalinference #SEM journals.sagepub.com/doi/10.1177/...

"Many widely used models amount to an elaborate means of making up numbers—but once a number has been produced, it tends to be taken seriously and its source (the model) is rarely examined carefully." Ouch. doi.org/10.1007/s000...

What is the best way to create a blog for #rstats content? Is it quarto blogs or are there better ways to do it?

This is really good.

A big problem I think

On the reproducibility crisis coming for epidemiology from 2008. I think still true today although probably a lot more good examples. Pre-specified primary endpoints + analysis plans, and corrections for multiple testing are not routinely used in observational studies. t.co/HPKYqhP1de

What are you enjoying on Netflix at the moment? I'm binging their series on the design of A/B tests, interesting to see how this is done outside of healthcare research. netflixtechblog.com/experimentat...

Could early phase trials use a Bayesian analysis? Priors centred on no effect with variance taken from systematic reviews like this one 👇. This might work like penalisation, shrinking extreme estimates and improving precision. Is anyone doing this?

New post: ‘Many analyses’ Some musings on studies that look at the impact of analysis choices on inference open.substack.com/pub/tpmorris...

Hi all! I'm looking for a short, accessible reading for 1st year UGs that covers: 1) contributors to the replication crisis(e.g., p-hacking, publication bias, researcher degrees of freedom) 2) initiatives to address it (e.g., preregistration, sharing data and code, registered reports). Any tips?

Is this inevitable? Ask me at 5pm today

Introducing PowerLMM.js! A new tool for power analysis of longitudinal linear mixed-effects models (LMMs) – with support for missing data, plus non-inferiority and equivalence tests. powerlmmjs.rpsychologist.com Would really appreciate your feedback as I refine this app! Details below 🧵👇

My collection of co-authored academic papers if I hadn't spent time scrolling bluesky:

Open source software is a can be vastly more impactful and important output than publications. Unis and funders place more importance on publications. This leads to unproductive, and uncollaberative (with regard to code) work with individuals having their own artisnal analysis pipelines.

Excellent thread. Open source maintenance is work. My repos might make look worse than what @seabbs.bsky.social shows here. And the is often (largely) unpaid, though we do by now have citation mechanisms. We can do better. We need to do better. CCing @sovereign.tech, a force for good here.

Never going to give correcting confidence interval interpretations up Never going to let statisticians down Never going to run around and turn Bayesian

Great read. And I haven't been rickrolled since, i dunno, maybe 2009 ;).

Also TIL you can make an Rmarkdown file output to github_document which will display in github. This allows you to create a 'readme.md' from a 'readme.rmd' file. This repo is a neat example.

This is wild: TIL you can splice the mapping in ggplot2: my_mapping <- aes(x = foo, y = bar) aes(colour = qux, !!!my_mapping) #> Aesthetic mapping: #> * `x` -> `foo` #> * `y` -> `bar` #> * `colour` -> `qux`

What is everyone’s preferred method of capturing your custom {ggplot2} theme - tiny #rstats package 📦 on GitHub or some type of local script?

I might just about manage skinny jeaned linear regression, given enough flat whites.

No.

Trying to find literature on multivariate prediction modelling (many outcomes) and almost every paper I can find has confused multivariate with multivariable. Anyone have any examples (real multivatiate analysis)?

Thank you! In more recent training materials like masalmon.eu/talks/2024-0... I also list resources. 🙂