naming your biology plagiarism machine after the most famous case of biology plagiarism is a very funny bit I must admit

I think policies that support editors to desk reject more papers could free up capacity Dichotomise continuous variable - reject Stepwise variable selection - reject Table 2 fallacy - reject Etc If widely applied, this could drive up standards

I’ve been posting and talking about mode effects a lot recently, but if you still don’t know what mode effects are and/or how to deal with them in your analyses, now is your chance to learn! Register to attend our FREE online event on 30-Apr👇🏻 @rjsilverwood.bsky.social @clscohorts.bsky.social

Science is a cumulative and collaborative process that requires human connection and interaction. I find it interesting that the 'lone genius' idea still gets romanticised in academia.

Recording now available if you've missed it:

TODAY at 3pm UK time! Register here: ucl.zoom.us/meeting/regi...

Replicated finding ≠ true* Unreplicated finding ≠ false** *because internal validity biases can and do replicate **because external validity refers to a specific population - the size/presence of an effect in a particular study is dependent on effect modifiers in the original sample/population

To gain false confidence in your results? The awkward fact is that structural biases replicate. The HRT and CVD studies of the 1980s all replicated the same horrifically biased result that HRT protects against CVD risk. This is a key argument for triangulation.

Late to the conversation, but I'm kind of unbothered by this lack of replication. If the study is not internally valid to begin with, then being able to replicate it only means replicating systematic biases. And something tells me this is likely to be a bigger problem in the literature.

economists be like "our worst quality is that we're very good at this."

Basically: academia is so hostile to a healthy personal life that the only way to have a romantic relationship and/or parent children is to bond with another academic?

Flying halfway round the world to spend 15 minutes failing to show 47 slides to four people, one of whom will respond by talking off topic at length while everybody sits there hoping someone else will tell him to stop, is a rational use of time and resources

It’s really quite shocking Wellcome didn’t have internal policy/feedback systems that prevent burning considerable resources on pseudo-science…

Our preprint is now out! 📄 We looked at whether five flexible IPTW / propensity score weighting approaches can be used for count exposures, and whether they can be combined with multiple imputation. We also illustrate the methods using BCS70 data. Link: doi.org/10.48550/arXiv.2603.23726

I love the internet

I do not want to use AI to write or think. Can someone train it to handle annoying tasks like formatting journal submissions instead of trying to get us all to use it for things that take critical thought and expert knowledge?

oil prices are about to make polyester a noble fiber

A very warm welcome to Bluesky to @helena-koerber.bsky.social! 🥳 Helena is a very talented early-career researcher with interests in research methods, surveys, and social attitudes. Please follow her and suggest the best accounts/feeds for her to follow!😊

Why do so many PhD students to this day get taught to "find the gap in the literature"? It has nothing to do with understanding what questions *need* answering.

Alternately the conclusions could be phrased "We spent many hours and tax payers money on a project to which we have at best made a very minor scientific contribution that better research is needed." Note the word "better" rather than "more".

Risky click

if you get this reference to early-'20s twitter, log off, touch grass, and burn your copy of The Book of Why

Bonus points: who are the compliers to a genetic instrument?

Another week in which I ask myself who are the compliers in a complier average causal effect and why are they our target population

Meta-analyses in nutrition may pool estimates for completely incomparable effects without realising it, just because of whether/how energy intake is adjusted. @nataliaortega.bsky.social will explain why, share results from a meta-scientific review & offer recommendations for ways forward 👇

📢 Researchers at @ucl.ac.uk propose new approach to improve quality of population data research by looking at data linked from the Millennium Cohort Study to the UK National Pupil Database (NPD). Read the full paper on the @ijpds.bsky.social journal website 👉 buff.ly/HJLkFkg

If you missed this seminar, you can: 📹 watch the recording here: youtu.be/n5HHg4RJQns?... 📖 read the (brand new!) paper here: doi.org/10.23889/ijp... cc @rjsilverwood.bsky.social @clscohorts.bsky.social @sriucl.bsky.social

This was very neat - paper here ijpds.org/article/view...

If you want to hear more about this research by @rjsilverwood.bsky.social, join our seminar starting in 30min! Get Zoom link here: ucl.zoom.us/meeting/regi...

Administrative data are commonly used in research but they often have limited information on confounders, making analyses prone to residual confounding. Join us to hear from @rjsilverwood.bsky.social about a multiple imputation-based approach to address this using linked cohort data. Link below👇

Just a few hours to submit your abstract to #SSMEuroEpi2026! Don't miss your chance to come to London for the first European Congress of Epi in six years - joint with the 70th @socsocmed.bsky.social! Conference & submission website: lnkd.in/edqNJi8u Location: QMUL, London, UK Dates: 9-11 Sep 2026

This recent RCT of an "AI stethoscope" claims the technology "shows promise" for diagnosing cardiovascular conditions. It does not. It is a textbook example of the risks of conducting unprincipled 'per protocol analyses'. Once again, peer review at a major medical journal has failed. 🧵 1/

When talking to people about research questions, I find people often make one of two mistakes: 1) They think that a causal question implies a causal answer 2) Or they seem to infer the question based on the methods/answer rather than the other way around

CAUSAL GRAPHS ➡️ HAPPINESS. 😍 Agree? & Work in/around Berlin? 🐻 Sign up for the 2026 Applied Causal Graphs Workshop! 🌟 Deadline Feb 28th ⬇️ applied-causal-graphs.de This 3rd edition is hosted by the great folks at Uni Potsdam 💡

Administrative data are commonly used in research but they often have limited information on confounders, making analyses prone to residual confounding. Join us to hear from @rjsilverwood.bsky.social about a multiple imputation-based approach to address this using linked cohort data. Link below👇

Really looking forward to sharing our work on understanding mode effects with DAGs! Join in person at LSHTM or online on 26 February 👇 cc @rjsilverwood.bsky.social @pwgtennant.bsky.social @clscohorts.bsky.social @sriucl.bsky.social

200+ attendees joined us for this seminar with Mats Stensrud today. If you missed it, you can watch the recording here: youtu.be/KL4eFZcw-sg

How can LLMs ever be compatible with reproducible science when they fail at reproducing their own response twice in a row?

A PhD candidate at Harvard Nutrition steps forward, head bowed. Walter Willett, dressed in full regalia, solemnly reaches into two fishbowls. One is full of slips of paper with nutrients/foods on them. The other, diseases. The random pair he withdraws is their dissertation topic. *Trumpets sound*

Neyman is my hero

Join us today/tonight to discuss the need for and challenges of conducting meta-scientific reviews! FREE for @societyforepi.bsky.social members!

“Causality is hard to let’s pretend we’re not studying it and definitely let’s not use all the tools designed to estimate causal effects”

Why am I so good at coming up with distracting exciting new research ideas when I have to focus on finishing something before a deadline??? Send help.

Die, Dichotomy. www.linkedin.com/pulse/die-di... A brief post encouraging you to read the original paper with @erik-van-zwet.bsky.social and @f2harrell.bsky.social . onlinelibrary.wiley.com/doi/epdf/10....

It is ridiculous, that they think the solution to simplistic quality assessments is adding more metrics. Here is my metascientist approach to see if someone is any good: read their work.

When it comes to dividing distributions into four equal parts may the fourth be with you and give no quarter to those who refer to them as quartiles.

New preprint 🥳 The city of Leipzig conducts large-scale surveys of adolescents. In 2023, the gender gap in life satisfaction has significantly widened, with girls declining more steeply than boys. What's up with that?> (work with @rmcelreath.bsky.social and @gregork.bsky.social)

Yeah. I see a lot of people saying "DAGs can't have loops, therefore I need a causal loop diagram." Which demonstrates a misunderstanding of both what a DAG and causal loop diagram does.

Yeah, that was the big challenge with this paper. Reviewers who knew DAGs were like “nonsense, everyone knows bidirectional effects dont exist” and those who didn’t know DAGs were like “nonsense, everyone knows bidirectional effects do exist” 🤦🏼‍♀️

"bidirectional causation" sounds like you are ready to embrace your causal aims, but not ready to let go of your old research questions