It's important to emphasize that "realistic-looking" data does *not* mean "realistic" data – especially high-dimensional data (unfortunately that post doesn't warn against this).
If one had an algorithm that generated realistic data for a given inference problem, it would mean that that inference problem had been solved. So: for educational purposes, why not. But for validation-like purposes, use with uttermost caution and at your own peril.
A #bayesian blogpost, by two of my undergraduate students! It's their report on their learning Bayesian modeling by applying it to my lab's data.
https://alexholcombe.github.io/brms_psychometric_variableGuessRate_lapseRate/DenisonBlog.html
Summary: we learned to use brms. But had trouble when we added more than one or two factors to the model. Little idea why; haven't had time to tinker much with that.
I'm explaining Hamiltonian Monte Carlo in my grad-level stats class tomorrow, so I put together this animation illustrating HMC in one dimension. I find it very soothing.
Got the physical copies of my Cambridge element in the mail. A reminder that the whole book is free to download until the end of February: https://doi.org/10.1017/9781009210171
GO GO GO! We're live next week in 40 George Square, #Edinburgh
Come and hear Isabella and Jordan talk about #Bayesian #Statistics in #RStats
Details: http://edinbr.org/edinbr/2025/02/17/February-meeting.html
I got an email from the author promoting this benchmark comparison of #Julialang + StanBlocks + #Enzyme vs #Stan runtimes.
StanBlocks is a macro package for Julia that mimics the structure of a Stan program. This is the first I've heard about it.
A considerable number of these models are faster in Julia than Stan, maybe even most of them.
Episode 120 is Live!
https://learnbayesstats.com/episodes/8f372809-3905-4110-8e1b-2f5ca1f95b33
In this LBS episode, Alexandre Andorra, Liza Semenova, and Chris Wymant explore the intersection of epidemiology, Bayesian statistics, and data science!
A #Bayesian, #DualProcessTheory inspired, multinomial approach to moral dilemmas found the
- “do no harm” impulse predicted by class, but not reflection.
- “some harm for greater good” responses predicted by reflection, not class.
New book on Bayesian inference and human cognition. I have always enjoyed material from Tom Griffiths and also from Josh Tenenbaum, and I expect this new collected chapters would also be excellent. If you want to explore more literature, the contributing authors of individual chapters are also wonderful.
"A p-value is an #estimate of p(Data | Null Hypothesis). " – not correct. A p-value is an estimate of
p(Data or other imagined data | Null Hypothesis)
so not even just of the actual data you have. Which is why p-values depend on your stopping rule (and do not satisfy the "likelihood principle"). In this regard, see Jeffreys's quote below.
Imagine you design an experiment this way: "I'll test 10 subjects, and in the meantime I apply for a grant. At the time the 10th subject is tested, I'll know my application's outcome. If the outcome is positive, I'll test 10 more subjects; if it isn't, I'll stop". Not an unrealistic situation.
With this stopping rule, your p-value will depend on the probability that you get the grant. This is not a joke.
"*What the use of P implies, therefore, is that a hypothesis that may be true may be rejected because it has not predicted observable results that have not occurred.* This seems a remarkable procedure. On the face of it the fact that such results have not occurred might more reasonably be taken as evidence for the law, not against it." – H. Jeffreys, "Theory of Probability" § VII.7.2 (emphasis in the original) <https://doi.org/10.1093/oso/9780198503682.001.0001>.
@paninid p-values, to a large extent, exist because calculating the posterior is computationally expensive. Not all fields use the .05 cutoff.
A p-value is an #estimate of p(Data | Null Hypothesis). If the two #hypotheses are equally likely and they are mutually exclusive and they are closed over the #hypothesis space, then this is the same as p(Hypothesis | Data).
Meaning, under certain assumption, the p-value does represent the actually probability of being wrong.
However, given modern computers, there is no reason that #Bayesian odds-ratios can't completely replace their usage and avoid the many many problems with p-values.
Do you use Probability Density Function (PDF) plots to show your results?
Discover how compressing PDFs impacts data interpretation in research!
Racquel Fygenson reveals surprising insights from our study on plot scaling. 
Video: https://www.youtube.com/watch?v=BEsD1RIIjCo&ab_channel=Dr.LacePadilla
Paper: https://vis.khoury.northeastern.edu/pubs/Fygenson2024ImpactVerticalScaling/
Hey! Please have a look at my lecture slides on 𝘉𝘢𝘺𝘦𝘴𝘪𝘢𝘯 𝘝𝘦𝘤𝘵𝘰𝘳 𝘈𝘶𝘵𝘰𝘳𝘦𝘨𝘳𝘦𝘴𝘴𝘪𝘰𝘯𝘴 with a forecasting application of the material implemented using my 𝗥 package 𝗯𝘀𝘃𝗮𝗿𝗦𝗜𝗚𝗡𝘀! 
https://bsvars.github.io/2024-10-be24-bsvarSIGNs
No sign restrictions are involved here! Just the VAR model with hierarchical prior and Bayesian forecasting. Juicy! 
The latest `brms` CRAN release added support for `priorsense` for easy prior and likelihood sensitivity analysis https://doi.org/10.1007/s11222-023-10366-5
```
> fit |>
powerscale_plot_dens(variable='b_doseg', help_text=FALSE) +
labs(x='Dose (g) coefficient', y=NULL)
> powerscale_sensitivity(fit, variable='b_doseg')
Sensitivity based on cjs_dist:
variable prior likelihood diagnosis
b_doseg 0.236 0.219 prior-data conflict
```
If you are interested in pursuing a #PhD at @statistics_UCL (Oct/25) in #Bayesian methodology and computational #statistics, co-supervised by Sam Livingstone and myself, feel free to reach out. A strong academic background and proven experience in these areas are essential.
New article published: https://doi.org/10.1038/s41562-024-01988-4
We have found that over 51 diverse languages, word-initial consonants are systematically longer than their counterparts in other positions.
While the study only analyzes observational data, we think this might be one of several cues for segmenting the acoustic stream into words - present in possibly most of the world's languages.
@EJWagenmakers However, I'm not Chat. I have an intuition: if the number of pancakes goes really out of hand, they are all equally bad, no one gave a probability of 0 any chance.
If this #bayesian #intuition is correct, it illustrates something interesting: being closer but completely wrong doesn't get rewarded. Better not be too narrow-minded when predicting.
And all #null #hypotheses are false :)
Small advertisement for my Ph.D. thesis and code, focused on #computervision for #robotics.
Using #julialang to implement #Bayesian inference algorithms for the 6D pose estimation of known objects in depth images.
TLDR: it works even with occlusions; needs <1sec on a GPU; does not need training; future research could focus on including color images / semantic information since SOA performs much better if color images are available.
doc: https://publications.rwth-aachen.de/record/985219
code: https://github.com/rwth-irt/BayesianPoseEstimation.jl
Today is the last day of the "Introduction to Bayesian Statistics for Demographic Science" @MPIDR
It was a great week with hands-on lab sessions with amazing instructors guiding the participants through the jungle of prior and posterior probability distributions to understand Bayesian inference. A big thanks to IMPRS-PHDS for funding this workshop & bringing Doug Leasure, Andrea Aparicio, & Edith Darin to sunny Rostock!
