Greg Atkinson

Happiness, humour, truth and, most crucially, courage are the qualities needed to repel the racist tide.
Because its architects offer whining, victimhood, lies and, most crucially, cowardice.

Histogram figure illustrating selection bias in estimating intervention effects. The central blue histogram shows the full population distribution of pre-scores (mean = 100, SD = 10). A green histogram (top left) shows a random sample of 1000 participants with similar mean (≈100) and SD (≈10), producing a raw intervention effect of ~8.7 and standardised mean difference (SMD) of ~0.86. A red histogram (top right) shows a conditionally sampled group (n = 1000, pre-score ≥ 110), with higher mean (≈115) and smaller SD (≈4.4). This sample yields a similar raw effect (~9) but an inflated SMD (~2.0) compared to the population. A corrected SMD brings it back closer to the true effect (~0.9). The figure illustrates how restricting samples to higher-performing individuals (e.g., “elite” athletes) compresses variance and artificially inflates standardised effect sizes, despite unbiased raw effects.

#SportScience, especially in elite sports, has an issue that can make effects seem more impressive than they really are... selection bias on the sample (explanation🧵 and a simulation pictured to illustrate below).

1/12

Why randomized controlled trials matter and the procedures that strengthen them Randomized controlled trials are a key tool to study cause and effect. Why do they matter and how do they work?

Read more about RCTs and why they’re so important: ourworldindata.org/randomized-c...

WHO statement on autism-related issues Follow @WHO for the latest updates

WHO statement on autism-related issues

The World Health Organization (WHO) emphasizes that there is currently no conclusive scientific evidence confirming a possible link between #autism and use of acetaminophen (also known as paracetamol) during pregnancy

Full statement bit.ly/47YsgwI

Good luck media fact checkers. You have my sympathy.

Find the full paper here 🔗 doi.org/10.1123/ijsn...

🚨LEAD FEATURE ARTICLE🚨 November 2025✨

We are very excited to announce our lead feature article for the November 2025 by Lolli et al titled ‘Understanding Treatment Response Heterogeneity Using Crossover Randomized Controlled Trials: A Primer for Exercise and Nutrition Scientists’.

Restating a prediction I made on twitter that university rankings will be a thing of the past 10 years from now, and we'll look back on the university heads that first led us away from them with a great respect and appreciation....1/

Nearly 900 fewer people injured since 20mph introduction in Wales Figures show a 25% reduction in the number of injuries on Wales' roads in the past 18 months.

Almost 900 fewer people have been injured on Welsh roads since the default speed limit was lowered from 30 to 20mph two years ago

Casualties on 20 to 30mph roads between July and September 2024 were the lowest for the three month-period since records began in 1979

👏 Evidence-led policy

So happy to see this one out! Thank you so much for your help @gregatki.bsky.social, your insight and analysis of the data were invaluable (as always!). Look forward to working with you again.

Basal Metabolic Requirements, Biomarkers of Cardiometabolic Health, and Anthropometric Measures of Obesity in Women and Men With Restricted Growth Conditions Population-specific thresholds have not been defined for the levels of adiposity and systemic biomarkers that predict chronic health risks in people with restricted growth conditions. Here, anthropom....

Congrats to @lucymerrell.bsky.social on her new paper, "Basal Metabolic Requirements, Biomarkers of Cardiometabolic Health, & Anthropometric Measures of Obesity in Women & Men With Restricted Growth Conditions". It was a pleasure to work on this project. onlinelibrary.wiley.com/doi/10.1002/...

Guidelines for Meta-analyses and Systematic Reviews in Urology Our guideline comprises points addressing the conduct and interpretation of systematic reviews and meta-analyses in urology. Application of the guideline would lead to a more considered interpretation...

This is big! Worked with great statisticians on guidelines for meta-analysis & systematic review. We discuss rationales for systematic review, evaluation & interpretation of heterogeneity, & common errors in network meta-analysis, funnel plots etc. www.europeanurology.com/article/S030...

It is first important to stress that pilot and feasibility studies are *not* about detecting intervention effects. They are about exploring key uncertainties ahead of the main trial 2/6

I seem to have been discussing pilot and feasibility studies a lot this week. One question often asked is how big should my pilot/feasibility study be? 1/6
#MethodologyMonday #124

✨ NOVEMBER 2025 ISSUE LINE-UP✨

How are we already on the last issue of 2025?! This year has flown by but lucky for you, the IJSNEM content keeps on delivering. Stay tuned for papers to feature 🔔

How RFK Jr.’s misguided science on mRNA vaccines is shaping policy − a vaccine expert examines the false claims Chaos at the CDC and the sharp move away from mRNA vaccines has public health experts alarmed.

Deborah Fuller, a microbiologist from UWash, wrote this article pointing out many of the false statements made by RFK in his senate hearing yesterday and explaining why they were false.

INSPECT-SR: a tool for assessing trustworthiness of randomised controlled trials The integrity of evidence synthesis is threatened by problematic randomised controlled trials (RCTs). These are RCTs where there are serious concerns about the trustworthiness of the data or findings....

www.medrxiv.org/content/10.1...

INSPECT-SR: A tool for assessing trustworthiness of randomised controlled trials.

A further problem is illustrated by the authors equating no treatment with placebo. For example, describing their second assumption, they state, ‘Implicit in this notation is that there is a single version of ‘no treatment’ that is consistently defined across all subjects in the RCT and external controls’ (p4). However, most clinical trials are add-on trials (Senn, 2002), even if not specifically identified as such. In placebo-controlled trials, one starts with standard of care and then subjects are either allocated to receive in addition the active treatment or a placebo to it. The statistical analysis plan for SUNFISH (F Hoffman-La-Roche Ltd, 2020) has a protocol summary as Appendix 1, which has this to say, ‘In addition to the study drug treatment, patients may continue to receive concomitant drug medication…’ (p186). This is very standard for clinical trials. It highlights that a key assumption in borrowing control data in this way is that there has been no evolution in the standard of care in the period since the trial.

One of the most common misunderstandings about the use and value of placebos in clinical RCTs, often made by both methodological experts and experienced trialists:

(from @stephensenn.bsky.social in academic.oup.com/jrsssa/advan...)

Issues in the determination of ‘responders’ and ‘non‐responders’ in physiological research New Findings What is the topic for this review? We discuss the dichotomization of continuous-level physiological measurements into ‘responders’ and ‘non-responders’ when interventions/treatments ar...

I view detection of HTE as informing future moderation studies for the identification of baseline predictors/subgroups of response for eventual clinical utility. The info in sect 6 of physoc.onlinelibrary.wiley.com/doi/10.1113/... may apply to your Q, but sect 7 highlights need for clear purposes.

Beneath the surface of the sum When genetic interactions matter and when they don't

I wrote about gene-gene interactions (epistasis) and the implications for heritability, trait definitions, natural selection, and therapeutic interventions. Biology is clearly full of causal interactions, so why don't we see them in the data? A 🧵:

Thanks for taking time to read the paper. We've been very worried about its length!
1) yes - general agreement although no formal comparison of underlying maths.
2) Our main aim was to cover more robust ways of detecting HTE than typically used. Shrinkage is prob. most useful for clinical relevance.

It was a rewarding labour of love to work on this paper with a fantastic team of authors, “Understanding Treatment Response Heterogeneity Using Crossover Randomized Controlled Trials: A Primer for Exercise and Nutrition Scientists”. @hk-ijsnem.bsky.social journals.humankinetics.com/view/journal...

Quantifying Running Economy in Amateur Runners: Evaluating VO₂ and Energy Cost with Model-based Normalization #sportsscience #sportsmedicine #exercisescience Quantifying Running Economy in Amateur Runners: Evaluating VO₂ and Energy Cost with Model-based Normalization

Another nail in the coffin for ratios - this one in the context of running economy, "Quantifying Running Economy in Amateur Runners: Evaluating VO2 and Energy Cost with Model-based Normalization", www.jssm.org/jssm-24-684....

6. Consideration 1 renders multiplicity likely. Consideration 2 renders it unlikely. IMHO, it’s good to explore the presence of additive/multiplicative effects rather than blanket assumption of the latter & use of %. Figure 2 and related text here is of interest: www-users.york.ac.uk/~mb55/msc/tr...

5/6. And 2, a possibly competing (to 1 above) expectation that treatment effects would be least for the athletes who are already world-class. Assuming multiplicity (and using % for effect size) naturally assumes that these superb athletes show the biggest (not smallest) absolute effects.

4/6. I think there are 2 considerations. 1, the pretty good chance that a ratio-level (bounded by zero) test measure (time, speed, power, etc) is heteroscedastic, meaning that general variability of the test measure is higher for higher test values in general.

3/6. Multiplicative treatment effects mean that the treatment effect is proportional to the level of ability. When this is present, data are usually log-transformed and effect magnitude expressed as a %

2/6. Additive treatment effects mean that the effect magnitude of the treatment is generally constant over the whole range of athletic abilities. Effects can then be reported in absolute terms in the units of measurement.

1/6. At present, I am thinking a lot about additive vs multiplicative treatment/intervention effects on tests of athletic performance, or athletic performance itself. I favour exploration rather than blanket assumptions.......

Lab samples

Pilot trials have a key role in preparing for definitive randomised trials, yet determining their sample size remains challenging.

This article provides guidance, methods, and tools for calculating and justifying sample sizes in randomised pilot trials
www.bmj.com/content/390/...