A big chunk of my time is spent harmonizing data sources so that they can be combined and compared. This is preventable. If you don't want to have to do this, make sure to spend time planning up front, so that you consistently name, code, and label variables across collections of the same items. 🙌

Of course this approach will out-compete careful, slow, rigorous, ethical clinical science.

At least as far as medicine goes, novice researchers unleashing machine learning algorithms on data they just happen to have lying around, hyping it all as ARTIFICIAL INTELLIGENCE and bewildering reviewer and editors has indeed led to more "impact", and this is a disaster that gen-AI will amplify.

Yippee ki yay! Headlines shout “growth is back!”. And pop the champagne corks! But, if growth is the answer (as many seem to believe it is), then remind me again what was the question? open.substack.com/pub/ganeshna...

Got this song stuck in my head for no reason

#3 - The focus on meat & butter is not consistent with evidence, which overwhelmingly points to sat fat raising LDL, and LDL being causal for heart disease. But they end up recommending 10% of calories from sat fat – new guideline, same as the old guideline.

#2 - Recommendations to reduce processed food, sugar, and refined carbs is totally fine, and also totally standard. It's too bad they miss the point on processing. Food isn't bad because it's processed. It's bad because manufacturers use processing to produce crap food.

In sum (!) I have 3 basic takeaways: #1 The focus on trials vs. observational evidence would be great if they actually followed it (see: artificial sweeteners). Observational evidence has been WAY over-interpreted, and is responsible for much conflicting advice (see: eggs).

The sole of my boot came off and I swallowed a bug.

It stole from beloved literature written by real humans to generate the list in the first place, it stole water and electricity and money in an age where those are often in short supply, and it's stealing our ability to have trust and confidence in the media at a time we need that more than ever

💡Our new research reveals the extensive labour of infant feeding for NZ mothers experiencing household food insecurity. Mothers showed remarkable determination & resilience - they stretched budgets, compromised their own diets & faced shame and stress when seeking support, affecting their wellbeing.

I don't use AI. I am old school and use my brain.

Do not cite an academic paper unless you’ve read it

If you’re working on clinical prediction models, whether regression or machine learning methods then TRIPOD+AI is for you Better reporting → better science → safer, more trustworthy clinical AI. Read the full guidance here --> www.bmj.com/content/385/...

So much human potential has been squandered thanks to men like him being promoted and given power, instead of genuinely talented and insightful people.

One underrated advantage of Zoom webinars is that they enable you to learn something about statistics while eating ice cream.

Save the Children NZ have actually talked to children themselves about online safety - if the focus was genuinely on children’s safety we’d be elevating their voices and idea of solutions. www.savethechildren.org.nz/assets/Files...

if you need some help understanding or explaining why this is terrible, start here: newsletter.sachajudd.com/archive/id-p...

This is terrible policy that likely forces all users to identify themselves to online services, and fails to recognise the reality of the world in which modern teenagers exist and communicate. It is a knee jerk policy response to a poorly understood issue defined by people who don't talk to kids

The fact that many of the cohorts used for nutritional epidemiology are large, and the studies tend to run for a long time, does not mean you can draw causal conclusions. I know this is boring! But it's a big reason we believe stupid stuff about food. www.nytimes.com/2025/11/04/w...

NEW PAPER The use of explainable AI in healthcare evaluated using the well known Explain, Predict and Describe taxonomy by Galit Shmueli link.springer.com/article/10.1...

This year explaining regression to the mean took forever but I’m sure next year will be better.

Fab lineup of free events and experiences for residents & visitors to enjoy, take part in across te pokapū tāone/city centre 😍🌆 this weekend from the Karanga-a-hape ridge to the shores of Te Waitematā. Music, dancing, looking, relaxing, sports, culture, plus a 🎄and a guy 🎅🏼#citycentre #loveablecity

A great time yesterday catching up with my former PhD supervisor, Jill 😀 from my early MSc days just starting in academic research, Jill taught me the importance of conducting robust studies & really understanding the 'why' - values which I've carried since & would like to pass on to future students

Have had a brilliant couple of days in Melbourne at an Adaptive Platform Trials workshop #ACTA25 & the exceptional Aus Trials Methodology Conference #AusTriM25. Best part was catching up with past PhD students, both succeeding in their careers 🙌😊

Public trust in science is damaged by false certainty in the face of uncertain evidence. The Lancet does the scientific community and the public no favours by presenting narrow perspectives as the scientific consensus. 🧵 3/4

Very happy to be in the wonderful city of Melbourne for the Australian Clinical Trials Alliance symposium #ACTA25. Even though I'm not Australian, I'm hoping to get a deeper understanding of this important area of research. There was also an urn of Solo at lunch #TheLuckyCountry.

You do not need to see the results in order to write ~2/3 of a quant. research paper in medicine/health.

📢 The RISE Study is recruiting! We’re exploring Australia’s readiness to invest in early childhood food security & federally funded food vouchers. Are you a public servant, maternal & child health or food security practitioner who wants to have a say? ⬇️ researchsurveys.deakin.edu.au/jfe/form/SV_...

We got a Marsden grant! Yay! 🎉 Me and my little business named as an AI 🤗 Such a great team of researchers to work with on a cool study about screens and sleep in teens 📱🛌

A list of all successful 2025 #MarsdenFund projects can be found below. Congratulations to all of the funded researchers! www.royalsociety.org.nz/what-we-do/f...

🌟 Our new publication sheds light on infant nutrition & food insecurity: nutrient intakes were similar across food secure & food insecure households. Severely food insecure infants had slightly ⬆️ free and added sugars intakes & BMI z-scores but no difference in weight status doi.org/10.1017/S136...

If you’ve noticed you’re not as strong as you used to be — it might not be all in your head. Next week, Dr Lara Vlietstra joins us on the #NESIPodcast to unpack how muscle loss starts sooner than you think — and how midlife is the perfect time to build lasting strength. Full episode 🗓️ Nov 3rd

We wrote an article explaining why you shouldn't put several variables into a regression model and report which are statistically significant - even as exploratory research. bmjmedicine.bmj.com/content/4/1/.... How did we do?

On the road a day late! Good thing too, if I’d left yesterday my car might have been dinged up by trees or loose trucks.

Making research methods work accessible and usable for researchers is a challenge! Great seminar discussing how to bring causal inference back into the frame.

Your regular reminder that unions represent PEOPLE. They are simply the voice of the PEOPLE they represent. Nothing more. Nothing less. Attack them and you’re attacking the people they represent. It’s pretty simple. #nzpol

The world has entered a new reality. Global heating will soon exceed 1.5°C, putting humanity in the danger zone where multiple climate tipping points pose catastrophic risks to billions of people. #ClimateBreakdown is real, it’s happening now. www.youtube.com/watch?v=5BVj...

“Living car-free is the most impactful behavior by far in terms of reducing emissions.”

To ask a pakeha majority to remove representation of an indigenous minority is ongoing white supremacist colonisation. #nzpol

Excellent new editorial and guideline on interpreting p values and interval estimates bjsm.bmj.com/content/earl...

Statistics has never been about coding, it is about experimental design, appropriate application and interpretation.

Just because an LLM can produce a report with various figures & charts doesn't mean it is good at statistics. Because good statistics is not about producing code. It's about deep knowledge of study design & conduct. In my opinion, 95% of all data science problems come from poor questions & design.

If it doesn’t survive the washing machine or the dish washer, maybe it was never meant to be part of your household in the first place.

Kind reminder: data driven variable selection (e.g. forward/stepwise/univariable screening) makes things *worse* for most analytical goals

A life update: this wound up paying off. I'm still applying to several jobs that are hopefully careers, but I've got my survival needs covered by a job at the airport that I'm apparently pretty good at. And instead of being in an office alone, which was driving me mad, I see 20k people a day.

4300 teens will be thrown off jobseekers as youth jobs disappear across the country. More stress, more homelessness, more domestic violence, more poverty, more crime, more prisoners, more profit for PPPs - National Party's back on track is right on track #nzpol

There is something especially f'ed up about university press offices that don't include direct links to research reports in their press-releases/marketing.

Mōrena. #tgif damp humido moods #citycentre 🏙️

In general I think it's hard to combat scientific misinformation when some of the best research is locked behind an academic paywall, while lots of nonsense gets published free for everyone to read in predatory journals.