Ryan Martin

@aaronggreen Nice picture explanation! My only minor complaint is that "forced" and "measure-dependent" aren't defined. The point is that possibility uses exactly the info in F, no more & no less; Bayes requires more, which creates opportunity for conflict, so further justification is needed

On possibilism vs probabilism Possibilistic first, bound your manifold, dataset, whatever THEN do your probabilism (or hybrid). Working on an infographic explainer... would love your feedback @statsmartin

@UQprofessional Thanks, Michael! I really like the idea in this paper, so I was glad to get some mostly positive feedback on it from peer-review

Congratulations to @statsmartin

@UQprofessional This was to distinguish true \theta and generic values \vartheta. Recently, I've been writing \Theta for the true unknown/uncertain parameter and \theta for generic values. I generally like the latter convention better

Today, I am using \vartheta in honor of @statsmartin.

@mic_caprio Thanks for the shout-out; our paper on CP-as-IP is linked below if interested. Lots of open problems remain, of course, and congrats to Michele et al for tackling the case where imprecision is induced by ambiguity in the labels arxiv.org/abs/2112.10234

Can #conformalprediction (CP) and Imprecise Probability (IP) go hand in hand? @statsmartin already told us so. In this new paper we show just how an IP approach to CP can be fruitful for classification problems with ambiguous ground truth. 1/3 arxiv.org/abs/2411.04852

@HarryDCrane You know I'm in!

The American Academy is long overdue. I will happily contribute any way I can to make this effort a success, starting with all lecture materials, and by making the resources at the Researchers.One organization available. Who else is with me? Let's make it happen.

@RdeHeide Congrats, great paper!

Missed it? The video of the SIPTA seminar of Ryan Martin is now available on our youtube channel! Check it out at youtu.be/1VL5P1YcXdo @statsmartin

Thanks for the invitation and the announcement. Looking forward to it!

@SrijanSengupta7 @NCSUStatistics Good to have a few in-house people give talks, young and old folks (wink wink). It's cheap and students should get the chance to see that their profs are as good as the outside speakers

3) Would you like to come and give a talk at @NCSUStatistics? Also, would you like to nominate someone (e.g., a junior colleague)? Please DM/email me We are a state uni with limited funding for speakers, but I'd do my best subject to the funding constraint.

Folks, I have been asked to chair the departmental seminar committee at @NCSUStatistics. I would greatly appreciate your input and perspectives on the following:

@1HereticalTruth Congrats, Adam! Coincidentally, my wife and I have been together for 24 years and married for 18 as of tomorrow

Married 17 years to my only wife today. Been together for 24. She has improved my life in uncountable ways. Thanks to God.

Thanks @1HereticalTruth and @PublishPeerless for having me on, I really enjoyed our conversation

@UQprofessional @predict_addict It's all just valid inferential models: arxiv.org/abs/2112.10234

@predict_addict Shafer is not ashamed to draw these connections.

TIL Glenn Shafer and Vladimir Vovk repackaged frequentist statistics as "conformal prediction," and it is now the hottest new thing in machine learning.

The special issue in IJAR devoted to ISIPTA'2021 is now complete! Thanks Jasper De Bock and Andrés Cano for the co-editing. See the work by @ArthurVanCamp @gertdecooman @statsmartin @jason_konek @Seb_des @benavoli @zaffama @RuobinGong and more at sciencedirect.com/journal/intern…

I've fallen behind on my ST790 course progress reports. The last few lectures have covered imprecise prob in statistical inference: Dempster-Shafer, generalized Bayes, and some new developments of mine. This week wraps up stat inference and moves to prediction & classification

(Late) ST790 Week 6 review: (a) Lower previsions & characterizations of coherence (b) Natural extension & generalized Bayes rule Week 7 preview: Comparison of Dempster's and gen Bayes rules (from @RuobinGong & @XiaoLiMeng1 2021); intro to imprecise prob in stat inference

I'm teaching a special topics course this fall semester (starts Aug 23) on Imprecise-Probabilistic Foundations of Statistics. Lecture videos, etc will be made available on my website, anyone interested is welcome to follow along/participate. @SIPTAtweets …rdpress-courses2223.wolfware.ncsu.edu/st-790-001-fal…

@NahanniFinanci1 Thanks, Sara, for the nice compliments. There's nothing wrong with keeping a calendar, I was just saying that not all academics plan out their week hour by hour as the original thread suggested. My memory is faulty too, so I only make as many plans as I can remember!

@statsmartin Ryan, you are a true mentor, and a patient/kind instructor, as well as great mastery in your subject! Thank you for the suggestion....I do have a calendar (although messy) due to my faulty memory.

In a nutshell why I will never look for a job in academia

@SrijanSengupta7 It doesn't take a Nobel to get in the "club"

True story. Here's one large-scale study on this: papers.ssrn.com/sol3/papers.cf…

@SrijanSengupta7 Big changes like this are out of our hands. What we can control is the example we set. I do actually read people's papers to form an opinion and I do write honest evaluations -- I've written negative evaluations of cases with papers published in top journals

@statsmartin I know this will not come to pass. But just as a thought experiment, what if tenure dossiers consisted of your work rather than a list of where you published it, and evaluators had to decide by, you know, *actually reading* your work instead of where it was published?

'Tis the season for tenure & promotion evaluations and I'm reminded of the backwards influence journals have. The quality of a journal is determined by the quality of the authors/papers it publishes, not vice versa.