Stanley E. Lazic

Probability | Prediction | PhD @Cambridge_Uni/@Kings_College | Book: |

[Stanley E. Lazic @StanLazic]

Great to be a part of the UKRN!

[UK Reproducibility Network @ukrepro.bsky.social @ukrepro]

We are pleased to have Dr Stan Lazic (@StanLazic, https://t.co/iJ9hNsy0OR) as a member of the UKRN Advisory Board. To find out more about the structure and membership of UKRN, visit https://t.co/Ekp08IMzNQ #UKRNSpotlight

[Stanley E. Lazic @StanLazic]

@mattwridley Excellent lecture.

[Stanley E. Lazic @StanLazic]

@FamedCelebrity @Cloudflare I've switched from LaTeX to Typst (https://t.co/akUvUd0fRX) for many documents. Simpler syntax and compile times are much faster.

[Stanley E. Lazic @StanLazic]

Yes, Mead was a statistician: https://t.co/2Kv21Dsgfp

The method makes sense for a lot of exploratory biological research. Psychologists are beholden to Popper and hypothesis-driven research, whereas biologists are more comfortable with "I wonder what happens when I do x?" experiments. The resource equation is perfect for this.

[Stanley E. Lazic @StanLazic]

There is a justification based on the error df and the t-statistic: https://t.co/41y1qz80k5
Mead's original argument is that if you are in exploratory mode, then you're wasting df estimating the error variance too precisely and you're better off asking another question (e.g. another dose level, time point, or another factor).

[Stanley E. Lazic @StanLazic]

@lakens @jfiksel1 I think it's ideal for exploratory studies where you don't have an estimate of the SD and/or an effect size in mind. Rather than pretend, the resource equation can be used to justify the sample size.

[Stanley E. Lazic @StanLazic]

References are mostly books. Festing and Altman as @jfiksel1 mentioned, plus Mead's earlier book "The Design of Experiments" (1988), and more recently, Mead, Gilmour & Mead "Statistical Principles of the Design of Experiments" (2012). This paper also talks about it: https://t.co/k1FFTR074F

The idea doesn't seem to have spread beyond animal studies.

[Stanley E. Lazic @StanLazic]

@cremieuxrecueil “Best” for whom? This implies value and moral judgments. Not sure empirical research has anything to say about this.

[Stanley E. Lazic @StanLazic]

@Lester_Domes @learnfromerror @david_colquhoun @vamrhein @dailyzad "Statistically detectable" is the best term I've come across and I've switched to using this -- but I don't know who first suggested it.

[Stanley E. Lazic @StanLazic]

@rlmcelreath It’s nice to have you back on X.

[Stanley E. Lazic @StanLazic]

@cremieuxrecueil "...people aged 21-40 who lived alone and had earnings between €1,100 and €2,600 per month".
What ever it is they studied, it wasn't *universal* basic income.

[Stanley E. Lazic @StanLazic]

@cavepaleo Hair? What about their arms!?

[Stanley E. Lazic @StanLazic]

Academic chatter has dropped to nearly zero on this platform. 🤔

[Stanley E. Lazic @StanLazic]

Some researchers are so thorough they p-hack their sensitivity analyses.

[Crémieux @cremieuxrecueil]

Robustness tests are supposed to show a study's results hold up no matter how you reasonably change the specification

But we live in a world with p-hacking, and people p-hack robustness tests

Compared to unshown robustness tests (blue), what we get is suspiciously significant!

[Stanley E. Lazic @StanLazic]

@cremieuxrecueil “Autism-like behaviour” is induced in rodents by maternal immune activation (e.g. by Poly(I:C)). It’s a standard way to create rodent models of autism. I wouldn’t be so dismissive of the immune connection.

[Stanley E. Lazic @StanLazic]

@paulg Dilution. As moral obligation extends in space and time, the number of people included in your moral reckoning grows exponentially. Your ability to help is fixed, however. If you try helping everyone, you’ll end up helping no one.

[Stanley E. Lazic @StanLazic]

https://t.co/p2WBNTqRxe

[Stanley E. Lazic @StanLazic]

I've been working on this for a couple of years but it still feels incomplete. If you know of other papers discussing a similar idea, I would be grateful for the reference. And comments are welcome.

[link to preprint in reply]

[Stanley E. Lazic @StanLazic]

@paulg @Sam_Dumitriu Sentencing (punishment) should be a function of (1) the past number of convictions, (2) their severity, and (3) the current offence.

[Stanley E. Lazic @StanLazic]

😀

[Claas Voelcker @c_voelcker]

I genuinely can’t tell if this is a joke…