Lightnews — Scholar-powered news

LightNews

Vilgot Huhn

@vilgothuhn.bsky.social

260 followers 540 following 620 posts

Confused PhD student in psychology at Karolinska Institutet, Stockholm. GAD, ICBT, mechanisms of change. Organizing the ReproducibiliTea JC at KI. Website: https://vilgot-huhn.github.io/mywebsite/ Personal blog at unconfusion.substack.com

vilgot-huhn.github.io

Posts Media Videos Starter Packs

Vilgot Huhn @vilgothuhn.bsky.social · 3h

I generally don’t know what to make of the AI field but I definitely find the technology, especially image and video generation, very surprising. If someone where to describe the technical details beforehand and ask me guess whether it would work I would have said ”no way”.

Reposted by Vilgot Huhn

samuel mehr @mehr.nz · 5h

Notably this problem is also solved 100% without any AI and instead, at zero cost, by journals adopting format-agnostic initial submissions (as many journals have already done)

Michael Eisen @mbeisen.bsky.social · 5h

Finally, someone has solved a real problem with AI! No more having to take a paper in the format for a journal that rejected you, and reformat it for a new journal. Well done!! formatmypaper.com

Vilgot Huhn @vilgothuhn.bsky.social · 5h

My friend @mkgood.bsky.social did so in the supplemental materials for one of his papers. That was the last time I saw it.

Vilgot Huhn @vilgothuhn.bsky.social · 15h

You know I think you should have the decency to be a bit uncomfortable with the moral conundrums we face in life, that’s all.

Vilgot Huhn @vilgothuhn.bsky.social · 23h

What do you mean?

Vilgot Huhn @vilgothuhn.bsky.social · 1d

Though to be a bit more humble I guess that reaction is itself "not engaging deeply in the literature".

Vilgot Huhn @vilgothuhn.bsky.social · 1d

We sometimes complain about the lack of ambitious theory building but on the other hand when I come across it I often get a feeling that the authors are moving too fast, not deeply engaging in the nuances of the literature, prematurely leaping ahead.

Vilgot Huhn @vilgothuhn.bsky.social · 1d

*perfect line. damn autocorrect

Vilgot Huhn @vilgothuhn.bsky.social · 1d

I think your point here works well with this wibbly wobbly table analogy that I just made (maybe?)
Like the fact that the corners of the plane get wobbly doesn't mean we get worse predictions for the flattened cloud of datapoints. (Just based on visual intuition.)

bsky.app/profile/vilg...

Vilgot Huhn @vilgothuhn.bsky.social · 1d

I think this works as an intuitive explanation for multicolinearity, right? When the cloud is flat the plane is attached in a very stable way. The table has many legs. But when we flatten that cloud with an association between X2 and X2 the table becomes "wobbly" orthogonal to that axis.

Vilgot Huhn @vilgothuhn.bsky.social · 1d

*X1 and X2

Vilgot Huhn @vilgothuhn.bsky.social · 1d

I think this works as an intuitive explanation for multicolinearity, right? When the cloud is flat the plane is attached in a very stable way. The table has many legs. But when we flatten that cloud with an association between X2 and X2 the table becomes "wobbly" orthogonal to that axis.

Vilgot Huhn @vilgothuhn.bsky.social · 1d

Personally I notice I’m spending more and more time here and less and less on the other place. On the other hand I think any imitation of the hellsite is itself the hellsite. (But there are multiple circles of hell)

Vilgot Huhn @vilgothuhn.bsky.social · 1d

I find it interesting that the idea of a robot rebellion is exactly as old as the word ”robot” (from the 1920 play Rossumovi Univerzální Roboti)

en.wikipedia.org/wiki/R.U.R.

R.U.R. - Wikipedia

en.wikipedia.org

Vilgot Huhn @vilgothuhn.bsky.social · 4d

spuriouser and spuriouser

Vilgot Huhn @vilgothuhn.bsky.social · 4d

Still better design that catholicism which you can just cheese at the final level if you use deathbed confession

Vilgot Huhn @vilgothuhn.bsky.social · 4d

Yeah tbh people said it would be hard but as long as you apply effortless effort (basically don’t attack and dodge all the time) it’s gg. Heard people talk striving would be incompatible with winning and only lead to more suffering but imo they just need to git gud at zen.

Vilgot Huhn @vilgothuhn.bsky.social · 5d

Here’s the idea in words:
A regression with two predictors is a plane through a cube. If we look at the cube from the top and x1 and x2 are strongly correlated we’ll see almost a percent line. Since that’s basically a line the plane would be free to wibble-wobble on that ”axis”.

Vilgot Huhn @vilgothuhn.bsky.social · 5d

So I thought of a visualization on the train for how multicolinearity works (maybe) but now I don’t have any computer access do make it/check it and I have clinical work all day. Pain.

Vilgot Huhn @vilgothuhn.bsky.social · 6d

Saw someone call the best model notation:

yᵢ ~ N(μᵢ, σ²)
μᵢ = a + bxᵢ

”Bayesian model notation” and there sure isn’t anything Bayesian about it (other than culturally)

Vilgot Huhn @vilgothuhn.bsky.social · 6d

The dream of the 2010s is alive in Portland 🎶

Vilgot Huhn @vilgothuhn.bsky.social · 8d

my new rule is that you’re not allowed to be interested in meta-science unless you’re also interested in science sorry.

Vilgot Huhn @vilgothuhn.bsky.social · 8d

that is a very clear answer i have no further questions

Vilgot Huhn @vilgothuhn.bsky.social · 8d

they are also in sweden. took me like twenty minutes on their buggy app to get the damn drop-box to open for me.

Vilgot Huhn @vilgothuhn.bsky.social · 8d

he's already worn a purple suit in wonka what more do you want?

Vilgot Huhn @vilgothuhn.bsky.social · 8d

In this context I think the problem is that even if you can explain the logic behind why "this is a Cohen's d" it's not necessarily comparable to other estimates that other call "a Cohen's d". The other problem is that it's often unclear in papers, at least to me, what has been done.