The defenders of British Values, having completed their work assaulting police officers, have returned home, leaving London a Dystopian Hellscape once again

Why does everyone who picks up the mic at this conference feel they have to mention this talk, whether grudgingly or positively??

Miguel Hernán is presenting at EFSPI and giving a frankly embarrassing rant about (addendum-style) estimands. While lots of people are constructively using it to complement causal inference, he is sticking to basic & straw-man gripes that are easily addressed by thinking or listening to others.

Interesting discussion with collaborators about whether it’s acceptable to say anything negative about PPI. They showed a quote that seems like a huge “therefore” lurch! 1/ researchinvolvement.biomedcentral.com/articles/10....

Question: In a paper, is it ok or distasteful to describe your own earlier paper as “seminal”/“pivotal”? I think so, and think it will ultimately lose readers… Follow-up question: does it matter if seminal/pivotal is actually an accurate description? In this case it very much is not 😜

I already think John Baker will be impressed. Can you help my friend out and improve on it?

After a month I can’t remember what my own code does by reading the source (well at least it takes a while, and I don’t think I’m the worst coder). My help files are for me. If they help someone else too, great.

Look, I’m just saying that “Users can find out what my package does by reading through the source code” is not a win-for-open-source. It’s just admitting you don’t care enough about your users to write decent help files.

ISCB46 has been really fun. So much that I’ve barely been on here. Admittedly “bRave little soldieRs” probably shouldn’t have made it into the slides but IN MY DEFENCE I MADE THEM WITH QUARTO 😇

It was logistic regression in scikit learn (scikit-learn.org/stable/modul...). Apparently, you could not fully turn off regularisation there for some time.

A few years ago there was a hoo-ha when someone discovered a well-used implementation of logistic regression in python used some regularization by default. Can anyone recall more details (or point me to a link because I'm clearly a bad googler)?

Simulation studies are widely used, but have historically lacked clear guidance around preregistration. In response to this gap, a team of researchers have developed a new template for preregistering simulation studies, now available on OSF. 🚀 Read our Q&A: www.cos.io/blog/intr...

Doing slides for ISCB with quarto (since they accept it I thought why not) and don’t really see what the fuss is about – with two slides it takes ~10s to compile. Looks ok. Genuine question: what is exciting about it? Is it code-related stuff?

Post just up: Is multiple imputation making up information? tldr: no. Includes a cheeky simulation study to demonstrate the point. open.substack.com/pub/tpmorris...

Alright Americans, I know you’re having a hard time with stuff at the moment, so here’s an olive branch: I’ll start spelling it randomization if you start pronouncing the letter Z as “zed” instead of “zee”. LMK

What is the right update frequency I’m losing my mind here

It’s a fine line to tread between using the same headshot for quarter of a century and updating it so often you appear vain

New rule: If you’re going to rant about something, you have to say what you think the thing is doing and what you think its proponents claim. Because I keep reading rants that reveal the problem is the ranter’s understanding.

Data in a csv format which we converted to a pdf which we printed which we then scanned which we embedded in a slide deck which we then encrypted which we lost the password to and cited as “available upon request”

Copilot is a real misnomer. The experience is closer to a kid sat in the front seat, pressing every button, turning the volume up and down, pulling the handbrake, beeping the horn, grabbing the steering wheel and incessantly asking "What's that".

Reckon I can submit this as a figure for a paper?

New post: “Marginal” and “conditional” are relative descriptions of an estimand

BMC Medical Research Methodology publishes some low quality papers but also some excellent ones. Saying that, their AEs (or reviewer suggestion system) are not doing well. I have rejected the last ~30* review invitations for not being within my expertise.

I was thinking about a bunch of OLD furniture my parents have and the ridiculously high quality. And then new processes emerge, maybe 60 percent as good but 20 percent of the cost. And perhaps that’s acceptable for a chair. I worry when those utilities are applied to ideas.

“UKRI to scale-up ‘radical’ change to peer review” [after screening Hunger Games and Battle Royale at an away day] www.researchprofessionalnews.com/rr-news-uk-r...

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

I'm wfh and someone is watching the first Jurassic Park in the other room. How am I supposed to work in these conditions?

For researchers with limited background in statistics, we have written this gentle introduction to simulation studies (with @timpmorris.bsky.social and other people from the STRATOS initiative): bmjopen.bmj.com/content/10/1...

The author of this paper, James Jackson, has been in touch to clarify and I'd like to reflect his explanation: Dr Jackson does not think evaluating coverage is new. Rather the idea is that evaluating coverage would improve on current practice in evaluation of synthetic data...

I'm working on an exploratory analysis of clinical trial data and am trying to see if I can make everyone (stats and clinical coauthors) mad by putting pvalues in Table 1 but nowhere else in the manuscript.

And another! This is amazing. I'm killing it.

Also:

Remember when someone published a paper that “invented” integration in the 90s? Well, for the 20s someone has apparently reinvented coverage as a performance measure. They even cited a certain tutorial paper that, if they had just read it… academic.oup.com/jssam/advanc...

We’re all thinking it all the time

When people say “statistical significance is not clinical significance”, what they’re actually saying is “statistical insignificance does not make me an insignificant person”

A good control arm is not a valid criticism of a trial. #CCR25

This illustrated guide to TMLE by @kathoffman.bsky.social is very good. I always struggle a bit with the nuts and bolts of TMLE so found it helpful to see the procedure broken down like this. I finally know what a “clever covariate” is! www.khstats.com/blog/tmle/tu...

We don’t need this advice so often on here but I want to remind us of some wisdom from twitter: - When someone is like that, tumbleweed is a good strategy - You can mute - Soft-block if you need to - Proper block (and pat yourself on the back when people whatsapp you screenshots) 😘

Attending #PSI2025 at Wembley stadium ⚽ this week? Don’t miss new research on a novel multiple imputation procedure for handling missing data in clinical trials targeting a treatment policy estimand by @suziecro.bsky.social James Roger & James Carpenter

Looking forward to this in todays Estimands and Missing data session @ #PSI2025 & being the estimands and missing data bridge between James Bells talk on composite estimand strategies with a continuous outcome & @timpmorris.bsky.social talk on how many multiple imputations you need!

Should have worn a suit today but then I didn’t know when I woke up that I’d find an apple big enough to obscure my face!

I still get post for "Michael Bubbles" from time to time, which is the name the previous owner of my house used on all his supermarket loyalty cards. (I used to use "Anne Omalie" once in a while.)

Richard Don-Don; pleased to meet you 🤝 A while ago me and the kids made up a bunch of names we found funny and they’ve become alter-egos for when we don’t want to give our real names. The kids go by Humanity The Puppy and William Bottoms.

What I have learned by running a biased survey on the meaning of scientific authorship: perhaps we should be more Orwellian! www.linkedin.com/pulse/all-an...

Starting to look like I might not be able to work at Harvard anymore due to recent funding cuts. If you know of any open statistical consulting positions that support remote work or are NYC-based, please reach out! 😅

Thanks @timpmorris.bsky.social et al - a wonderful course at UCL on Simulation - highly recommend. And huge congratulations to my colleague Ryan for completing this complex methodological review - will be so important for #DIAMOND going forward. 🎉👏🏼

Wait what? www.jclinepi.com/article/S089...

WHY SHOULD WE FANTASIZE ABOUT RECRUITING ALL THESE GREAT ACADEMICS TO CORK/IRELAND WHEN WE ARE ALREADY HERE, YOU ABSOLUTE FOOLS.