Jeremy Labrecque 🇨🇦

I always wonder if people are aware that people in the future can check their predictions

Not really. But I think it would have been cool to learn.

I quit civil engineering after one semester for the same reason!

The day of your PhD Defence The day has finally come: you get to defend your PhD thesis. Read what the day of your defence will look like.

At Dutch PhD defences you must have two people ("paranymphs") who literally stand behind you for protection.

If I'm not mistaken, I think in Finland you get an actual sword?

Frankly, releasing LLMs which can give what seems like expert advice (on any subject) without having actual subject matter experts rigorously test it should be considered gross negligence.

Does it work well for collaboration?

Had never heard of typst. What do you use it for?

The main problem here is the asymmetry. Nobody ever talks about drivers as entitled, for example, but if you step back and think about it, it is pretty wild that we built the single biggest structure in human history and that only people who spend thousands of dollars on private property can use it.

I should tag that this was work with @gelovennan.bsky.social and @jessekrijthe.bsky.social!

Especially not having it in the documentation, to me, seems like a huge oversight.

My take is that it's, let's say, suboptimal, when software that can only be used to estimate causal effects labels the output as a causal effect estimate (e.g. ATE, ATT) but says nothing in the output or anywhere in the documentation about causal assumptions required for that interpretation.

Causal clarity in statistical software Imagine running a simple regression in any statistical software of choice—but this time, you only get a point estimate of the regression coefficient. There

Should statistical software that estimates causal effects also tell you the causal assumptions under which that estimate can be interpreted as causal?

I don't know but my PhD student Maurice Korf has some thoughts (and software) to get the conversation going:

academic.oup.com/ije/article/...

#overthinkingagoodjoke

This is funny. But doesn’t the dad know that ‘not significant’ doesn’t imply the null is true?

100% this, null results are valuable and I don’t trust the people upthread saying that they aren’t!

I’m also baffled. How can you ask a question where the only valuable answer is a positive answer? If this is someone’s attitude to science it has a suspicious resemblance to fishing.

SAVE THE DATE: The 2026 IEA European Congress of Epidemiology and 70th @socsocmed.bsky.social Annual Conference will take place in London, UK on 8th-11th September 2026!

#EpiSky #EuroEpi2026

I do this with papers I really love. Recently it led to writing a paper together with someone I contacted in this way!

It’s not like there aren’t enough important and genuinely complex issues we face, but it feels like the current trend is to take non-issues with obvious solutions and do the opposite.

Oooohhh I do remember this!

I know what a GRE score is but I don't understand the reference...so....

So "epidemiologist" will always be a mixed bag of expertises that will do their best work when they work together.

Thanks for the support Peter but maybe I wouldn't go so far as to call it a joke. Maybe it's just a reminder that it would be really difficult to define a protected title such as epidemiologist. Very few epidemiologists are experts in both the substantive and methodological side.

Yeah, you're right it doesn't impact me much (except to tell students they can get an PhD in epidemiology with me but can't be an Epidemiologist which hasn't mattered yet). But I think it overvalues the medical aspect of being an epidemiologist and undervalues the social and methodological aspects.

Epidemiologist I meant. It's of course the VVE's prerogative to decide who they think is an epidemiologist. But I am disappointed by this requirement.

I did a bachelors in environmental science and human health. I've taken many courses on genetics, cell biology, toxicology and I'm told none of them count. I have no time to take 8 credits of 'health and disease'. Am I forever prevented from being an epidemiology?

In the Netherlands it is protected but you have to either have studied medicine or take an exam to show you know medical stuff. Therefore I am not an epidemiologist here.

Yeah. What I explained is what explains the non-representativeness they talk about. But it takes a while to get there. I found this paper after I was thinking about e*c interactions. It’s a very specific rabbit hole I went very deep into.

Similar thinking to parametric standardization where you fit separate regression models to exposed and unexposed and use each model to “impute” the outcome of the other group.

It’s originally an Angrist paper but more elegantly explained by onlinelibrary.wiley.com/doi/abs/10.1...

But if c1*c2 is on a backdoor path then it will always bias the exposure estimate. So it’s definitely an it depends situation but after thinking a bit I’d worry more about c*c interactions.