iiinova — where your digital self keeps evolving

Teague Henry @teaguerhenry.bsky.social · 15/01/2026

Hello all! I’m recruiting a postdoc to work with my lab and I on methods for analyzing intensive longitudinal timeseries of psychological phenomena, with particular focus on measurement and optimization of interventions! Start would be August 2026

jobs.virginia.edu

Research Associate in Psychology in Charlottesville, Virginia, United States of America | Research at University of Virginia

Apply for Research Associate in Psychology job with University of Virginia in Charlottesville, Virginia, United States of America. Research at University of Virginia

082

Teague Henry @teaguerhenry.bsky.social · 30/04/2025

Yes! I tell my class that regression coefficients are the "unique" relationships of a variable. If variables are not "unique", there is less information available to make the SEs "smaller." It's always either multicollinearity or normality assumptions that you've got to disabuse students of!

030

Teague Henry @teaguerhenry.bsky.social · 17/04/2025

Wait, you're telling me that a study in 20 children with ADHD showing an increased correlation between 2 arbitrarily defined brain regions doesn't represent a paradigm shift in how we could treat ADHD with respect to possible targeted pharmaceuticals?

020

Reposted by Teague Henry

Margaret Mitchell @mmitchell.bsky.social · 09/04/2025

The one simple trick to avoid > 75% of bad uses of AI that AI ethicists don't want you to know! Read a science fiction book and if your goal aligns with the bad guys in that book don't do it. -- Credit @stellaathena.bsky.social on the Bad Site.

Gif of police officer, dramatic high contrast lighting, looking confused and asking, "Are we the baddies?"

56918

Teague Henry @teaguerhenry.bsky.social · 08/04/2025

I mean, why do journals give due dates on RRs if not to encourage us to submit at the very last second?

010

Teague Henry @teaguerhenry.bsky.social · 26/03/2025

Me, moments before I unknowingly hit reply all on the faculty listserv: "We are currently clean on OPSEC"

030

Teague Henry @teaguerhenry.bsky.social · 25/03/2025

Homebrew warlock patron "The P-factor" that lets you summon and bind fiends by drawing path diagrams for multi-level SEMs. At level 17 you gain the ability to measure constructs ... without error!

040

Reposted by Teague Henry

Crystal Lewis @cghlewis.bsky.social · 23/03/2025

In 2018 I gave a presentation on research data management to a group of graduate students. Not surprisingly, most of the problems and solutions I discussed 7 years ago, are still the same problems and solutions I see today. #databs

0245

Reposted by Teague Henry

UVA Department of Psychology @uvapsychology.bsky.social · 18/03/2025

🧠 Applying to Psychology PhD programs? Join @UVA's virtual panel April 14 at 12:30PM! Current faculty & PhD students will share tips on choosing programs & creating competitive applications. All welcome, especially those from underrepresented groups. Register via QR code! #PsychPhD #GradSchool

055

Reposted by Teague Henry

dennyborsboom.bsky.social @dennyborsboom.bsky.social · 18/02/2025

This year I have the honour of serving the Psychometric Society as its President, and we have been working hard on the program for the International Meeting of the Psychometric Society (IMPS) psychometricsociety.org/imps-2025 You can submit an abstract here: imps2025.exordo.com

psychometricsociety.org

IMPS 2025 - Psychometric Society

July 15-18, 2025 (Short Courses July 14)

1147

Reposted by Teague Henry

Eiko Fried @eikofried.bsky.social · 13/02/2025

1/3 Tutorial on exploring ecological momentary assessment data is online at AMPPS, with: - Accessible ways to visualize data for better understanding - Models to get some first insights - Further reading boxes for more advanced topics - Reproducible pipeline you can run over your own data

615477

Reposted by Teague Henry

Jonas Haslbeck @jmbh.bsky.social · 13/02/2025

New preprint by Anja Ernst and me on Modeling Qualitative Between-Person Heterogeneity in Time-Series using Latent Class Vector Autoregressive Models (osf.io/preprints/ps...)

151

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

... to use human raters to reconcile any ambiguities. I think this is fascinating work, and a great example of how to use LLMs as a true tool for analyzing data (rather than just asking an LLM to analyze data). Joy is currently working on an extension to discover new topics, which is very exciting!

040

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

We found that ensembles of small LLMs tended to have reasonable (better than chance) performance at identifying topics. It wasn't perfect, but it is good enough to be a reasonable first pass through a large dataset. The idea is to first identify the cases that are easy to classify, then...

120

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

We compared the LLM topics to human raters topics. Now, importantly, Joy used an ensemble of LLMs (equivalent to having several human raters). Why? Well, we wanted for this method to be able to run locally, as there are a number of privacy issues with using consumer LLMs to analyze health data.

110

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

She applied this method to both a massive dataset of Reddit posts from eating disorder related forums, and a smaller dataset of free-text responses from patients with eating disorders (courtesy of @cherilev.bsky.social), and compared how well these LLMs could identify a prespecified set of topics.

120

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

So, Joy decided to use LLMs to query free-text data to extract "topics." For example, if a study collected responses regarding stressful events, then a topic might be a description like "work stressor" or "interpersonal stressor."

110

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

Enter LLMs! While I have a number of thoughts on LLMs that I won't get into here, for this project we conceptualized them as "tools that can comprehend free-text," with comprehension being akin to human reading comprehension (do LLMs comprehend? No, but they sure do seem like they do!)

110

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

Traditionally, this is done via teams of human raters. Researchers will put together groups of raters (usually grad students, possibly RAs) that will go through the free text responses and quantify them in some way. This takes an enormous amount of time and resources.

100

Teague Henry @teaguerhenry.bsky.social · 11/02/2025

Free-text responses are a wonderful way of collecting nuanced information about any number of psychological phenomena, but the issue is, of course, that they are free-text responses. To perform quantitative analysis on them, you need to convert them into some set of numbers.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

We originally thought about this project from an EMA perspective, trying to optimize for participant burden. So that’s where we are coming from!

000

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

To your second, I think this matters the most when only one/a couple questions per construct are asked. For a multi item scale, while I still want to see measurement errors of individual items, I don’t think the extra error induced by VAS would degrade wholistic scale performance.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Fortunately, it is an empirical question, and I would be happy with seeing that no, VAS doesn’t meaningfully degrade measurement! I just want to see the studies.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

To your first question: yes, but different things matter more or less for measurement quality. For example, I wouldn’t think that a slight change in anchor language needs a revalidation (it could though!)

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Found my new standard review comment

120

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

So yeah, the jury is still a bit out on the use of VAS scales. It seems like they can work well, but in light of the potential measurement properties, we as a field need to do more work validating their use for specific scales. I'm definitely not comfortable with just changing an existing scale!

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

@jmbh.bsky.social and colleagues have this excellent paper osf.io/preprints/ps... comparing Likert and VAS scales in EMA, showing that VAS scales give higher magnitude correlations with trait measures.

osf.io

OSF

110

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

There has been some great work recently examining VAS scales in empirical data. Leonie Cloos (first author), @omidvebrahimi.bsky.social (last) and colleagues' work empirically looking at VAS in EMA studies.

osf.io

OSF

120

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

We strongly suggest that, if a scale is going to be converted into a VAS, then it needs to be revalidated using the VAS responses. After validation, one can directly compare the measurement error of the differing response types to see if it has increased.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

What we are most concerned about is researchers taking a scale that has been validated with a given set of response options and changing it to a VAS version. If measurement error is related to number of response options, that means that scale validation is specific to the number of response options.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

So what does this mean practically? Are we saying that we should reject modernity (VAS) and return to tradition (Likert)? No (though personally I am a skeptic about the use a VAS scales generally)...

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

So, we ran some simulation studies and found that yeah, if measurement error increases with the number of categories, then there is going to be an optimal number of response categories for a given scale, and that increasing beyond that point is going to lead to issues.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

There is a big issue with the simulation studies of response category numbers: the ones we reviewed made the assumption that the number of response categories is independent of the measurement error of the scale. While from a technical side, this makes sense, this assumption is prima facie wrong.

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Now the literature on this is mixed: Empirical studies have shown that optimal numbers of response categories are anywhere between 7 and ~30. Simulation studies have shown that measurement quality increases with the number of response categories up to a point, then it plateaus. BUT....

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

the additional "variance" due to having more response categories is not variation on the underlying construct, but rather pure measurement noise. So, we wanted to see what happened to the measurement properties of psychometric scales when the number of response categories was increased.

110

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

The crux of the theoretical argument is that there must be a resolution limit to how well humans can distinguish between response categories in psychological scales. This is really important, because if this resolution limit is exceeded (by having too many response categories), then...

110

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Let's get let a little silly: What if we just asked people to rate their mood as a normally distributed variable? So, now, we have an infinite number of potential response options. Can a person be reasonably expected to distinguish between a mood of 1.2313213 and 1.2313214?

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Okay, so far so good, but now let's increase the number of possible responses, from 7 to 10. Is it easier to distinguish between a 1 and a 2 when there are 10 options vs 7? How about when there are 100 options? How about when there are 1000, or 1 mil options?

100

Teague Henry @teaguerhenry.bsky.social · 08/02/2025

Putting on our "human being thinking about thinking" hats for a minute: If I ask you to rate your mood on a 1-7 scale (real bad mood - real good mood), I think we would all agree that humans could distinguish between a "real bad mood" and a "bad mood." (i.e. we can distinguish between a 1 and a 2).

100

Reposted by Teague Henry

Dr. Cheri A Levinson @cherilev.bsky.social · 30/01/2025

Hiring alert!! Looking for a PhD level research scientist and a data manager! Come join us!

043

Teague Henry @teaguerhenry.bsky.social · 30/01/2025

Raising my hand in a committee meeting: “this is more of a sabotage attempt than a question…”

120

Teague Henry @teaguerhenry.bsky.social · 20/01/2025

My favorite journal going Open Access is not something I had on my 2025 list! Fantastic news!

010

Reposted by Teague Henry

wint @dril.bsky.social · 14/01/2025

They cant get my data if i am Fully Dissolved in a bath tub full of sodium hydroxide

324208372896