Compositional data (proportions that sum to 1) behave in ways standard models aren’t built for

I walk through why Dirichlet regression is often the right tool & what extra insight it gives, using a real example of eyetracking

#Dirichlet #r #brms #guide #eyetracking #ordbetareg #tutorial #reanalysis #substack

https://open.substack.com/pub/mzloteanu/p/where-were-you-looking-re-analysis?r=3b457w&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Where were you looking? Reanalysis of Satchell, Hall, and Jones (2025)

An illustration on analyzing composite probabilities using Dirichelt models

Figuring Stuff Out - Dr Mircea Zloteanu

#statstab #478 Stuck between Zero and One: Modelling Non-Count Proportions with Beta and Dirichlet Regression

Thoughts: A few resources for proportion data.

#Dirichlet #proportions #betareg #count #tutorial

https://methodsblog.com/2019/08/06/beta-dirichlet-regression_en/

Stuck between Zero and One: Modelling Non-Count Proportions with Beta and Dirichlet Regression

Post provided by JAMES WEEDON & BOB DOUMA Chinese translation provided by Zishen Wang 這篇博客文章也有中文版 Imagine the scene: you’re presenting your exciting research results at an important internation…

Methods Blog

#statstab #477 Simulating data for Dirichlet regression with varying estimates

Thoughts: Interesting thread about an underused model.

#Dirichlet #brms #poweranalysis #rstats #proportions #stan #forum

https://discourse.mc-stan.org/t/simulating-data-for-dirichlet-regression-with-varying-estimates/35325

Simulating data for Dirichlet regression with varying estimates

Hello everyone, The shortest formulation of my problem is the following: I’m trying to design a power analysis by simulation for a Dirichlet regression model using brms. How could we calculate the shape parameter alpha of the Dirichlet distribution by hand from a theoretical model in order to feed it to brms::rdirichlet(n, alpha) to simulate data with chosen effects? The goal is to tweak manually the coefficients of the model to simulate data with varying effect sizes. Now for the lengthier c...

The Stan Forums

#statstab #439 {ZOID}

Thoughts: "zero-and-one inflated Dirichlet regression (also known as trinomial mixture models) in a Bayesian framework (Stan)"
Useful for composite proportions.

#Dirichlet #bayesian #stan #r #rstats #proportions

https://noaa-nwfsc.github.io/zoid/index.html

Bayesian Zero-and-One Inflated Dirichlet Regression Modelling

Fits Dirichlet regression and zero-and-one inflated Dirichlet regression with Bayesian methods implemented in Stan. These models are sometimes referred to as trinomial mixture models; covariates and overdispersion can optionally be included.

#statstab #435 Guide to understanding the intuition behind the Dirichlet distribution

Thoughts: Useful for composite proportions, but take ages in brms.

#brms #Dirichlet #proportions #modelling #rstats #r #betareg

https://www.andrewheiss.com/blog/2023/09/18/understanding-dirichlet-beta-intuition/

Guide to understanding the intuition behind the Dirichlet distribution | Andrew Heiss

Learn about the Dirichlet distribution and explore how it’s just a fancier version of the Beta distribution

Andrew Heiss
Die #Trendprognose⁠n bei der #wienwahl waren wirklich schlecht, auch im Verhältnis zu den normalen #Umfragen. Hier ein Vergleich mit verschiedenen Abweichungsmaßen. Erste Datenspalte ist die #Wahrscheinlichkeit, dass eine #Stichprobe mit 600 gültigen Wählern netto noch schlechter ist, definiert als Werte mit geringerer #Dirichlet-#Wahrscheinlichkeitsdichte. Nächste Spalte dito, aber alles auf halbe Prozent gerundet. [1/2]

Wir wünschen allen Alumni der #GAUG einen schönen Alumni-Tag u.a. bei uns im MI im Auditorium Maximum ab 13:00 mit folgenden Programm:

* Grußworte, Neues aus der Fakultät und Alumni Chapter (Dekanin Prof. Dr. Anja Sturm, Prof. Dr. Ralf Meyer)
* Festvortrag von Prof. Dr. Ina Kersten, Zum Leben und Werk von Gustav Lejeune #Dirichlet, Mathematiker und Nachfolger von Carl Friedrich #Gauß mit musikalischer Begleitung
* Jubilarsfeier mit Verleihung von Ehrenurkunden
* Empfang mit Getränken und Kuchen

New large value estimates for Dirichlet polynomials

We prove new bounds for how often Dirichlet polynomials can take large values. This gives improved estimates for a Dirichlet polynomial of length $N$ taking values of size close to $N^{3/4}$, which is the critical situation for several estimates in analytic number theory connected to prime numbers and the Riemann zeta function. As a consequence, we deduce a zero density estimate $N(σ,T)\le T^{30(1-σ)/13+o(1)}$ and asymptotics for primes in short intervals of length $x^{17/30+o(1)}$.

arXiv.org

'Choosing the Number of Topics in LDA Models – A Monte Carlo Comparison of Selection Criteria', by Victor Bystrov, Viktoriia Naboka-Krell, Anna Staszewska-Bystrova, Peter Winker.

http://jmlr.org/papers/v25/23-0188.html

#topics #lda #dirichlet

Choosing the Number of Topics in LDA Models – A Monte Carlo Comparison of Selection Criteria

I knew of the Dirichlet function, but only as a funny example when learning about discontinuous function. Happy to remedy that today by stumbling across this brief video explaining the motivation behind it. It also increased my knowledge about the weirdness of infinities by showing proof that you can always find a rational between two irrationals, and vice versa.
https://www.youtube.com/watch?v=uRluS4KwXu8
#math #maths #mathVideos #Dirichlet #infinity
Dirichlet Invented this Function to Prove a Point

YouTube