@brunozumbo.bsky.social
Professor & Distinguished University Scholar, University of British Columbia. Mathematical models of mental test scores. Statistical methods for educational & behavioral science. https://ecps.educ.ubc.ca/bruno-d-zumbo/ Private account; all opinions my own
📌 Companion paper to recently published paper “Reliability as a theorem” that extends test theory into nonlinear and information‑geometric spaces. Unified framework for expectation, test theory, reliability, and item response theory.
doi.org/10.1177/0013...
📌 Reliability as a theorem: Reframed not as an assumption of classical test theory, but as a direct consequence of Hilbert space projection geometry; interpret true scores
Recently published in 𝘌𝘥𝘶𝘤𝘢𝘵𝘪𝘰𝘯𝘢𝘭 𝘢𝘯𝘥 𝘗𝘴𝘺𝘤𝘩𝘰𝘭𝘰𝘨𝘪𝘤𝘢𝘭 𝘔𝘦𝘢𝘴𝘶𝘳𝘦𝘮𝘦𝘯𝘵, Online first, open access 𝐟𝐮𝐥𝐥 𝐚𝐫𝐭𝐢𝐜𝐥𝐞: doi.org/10.1177/0013...
Last call for papers for the 10th integrated HPS conference: March 27 - March 29, 2025 in Pasadena, California. Please consider submitting and share widely. (Deadline: August 18)
#philsky, #philsci, #histsci, #histSTM
A couple of weeks ago I had the pleasure of talking to Carmelina Conterino at the HPS Podcast.
True-dat
Now all I can think about is how would one blow smoke up their own ass.
Sure, I can imagine it. It’s not as if I lack imagination. However, I wonder whether it’s even possible at this point in my life when I feel back pain just getting out of bed?
At last week's NCME conference, I did a workshop on Bayesian latent variable modeling with a focus on the blavaan package. We discussed factor analysis, item response models, and two-level structural equation models. Slides and code are at the first bullet:
Sometimes the best way to support academic freedom is to tell those who summon you that they have no jurisdiction
Free webinar for researchers and data analysts conducting survey research. Provides a description and demo of the impact of ignoring survey nonresponses on your conclusions — role of statistical models.
YouTube - youtu.be/2jS58Zs7eek (2:07:53 in length); PDF of lecture notes at tinyurl.com/y8zbdv5n
2/2 🧵Regarding the effects of measurement error on regression coefficients to understand why this approach is problematic, we distinguish between four related regression models described on page 417 of Shear, B. R., & Zumbo, B. D. (2013) doi.org/10.1177/0013...
1/2 🧵Regarding the effects of measurement error on regression coefficients, see Figure 2 and the section titled "Errors of Inference Due to Data–Model–Inference Mismatch" in Shear, B. R., & Zumbo, B. D. (2013) doi.org/10.1177/0013...
Congratulations! I look forward to reading it.
This insightful article by Brian Haig should go a long way toward clearing the muddy waters surroundiing EFA.
Haig B. D. (2005). Exploratory Factor Analysis, Theory Generation, and Scientific Method. Multivariate behavioral research, 40(3), 303–329. doi.org/10.1207/s153...
🧵 2/2 Key topics:
- historical analysis highlights explanation-focused;
- role of ethical and social values;
- measure-theoretic mental test theory;
- novel methodologies, ecological model of item and test performance;
- Third-generation DIF is not just about item bias.
🧵 1/2 New open access paper:
First soup-to-nuts description of my explanation-focused view of test validity.
A dialectic on validity: Explanation-focused and the many ways of being human. “International Journal of Assessment Tools in Education,” vol. 10, pp. 1-96. doi.org/10.21449/ija...
And maybe, just maybe, in some (many) measurement settings one cannot recalibrate or rescale the problems away.
Psychometric math washing may not be the answer. It may add (dare I say) innumerable additional problems and misinterpretations.
It’s almost as if there is a moral basis to evaluation and measurement; and talk of a moral consequentialist stance. And, maybe this moral turn coincided with a recent social turn in phil of science? And this moral and social turn in measurement has changed how we think about validity? Who da thunk?
… and what does it mean to “correct” a score in this context?
… about the measurement process, the construct, and psychometric characteristics such as measurement error and dimensionality of the item data.
I would focus on the interpretation and use of the “adjusted” scores, which also involves close conceptual analysis of what it purports to measure …
The “adjustment” or “correction” should be described the public technical reports or manuals.
The adjust methods typically involve some form of re-calibration, re-scaling, or adjusted scoring so your intuition is spot on. There are various classes of methods for this that make strong assumptions
My recent manuscript included: “Alan Richardson is solely responsible for any remaining errors.”
Keep an eye out for it.
AI-driven what? Have they no shame?
I just felt part of brain shut down in protest of cerebral abuse. And that was the part of my brain that was working on breakfast plans. Thanks!
I’m onto you. Nudge me closer to the edge of madness, why don’t you?
Soon you’ll find me wearing a tin foil hat selling techno-eval gizmos out of the trunk of my car to help track your pillars of excellence.
You’ll hear a voice in the distance … “So, youse want dat median wit or witout waits?
Typically, I should be able to get a sense, even the bare essentials, of a survey and evaluation system in less than half that time.
This is troubling.
I spent two hours searching for details about Explorance’s survey methods and for any evidence of their claims regarding the validity (in technical sense) of their tools to no avail.
E.g., lots of slogans and marketing but I cannot find anything to support their claims about “metrics that matter.”
If a pigeon’s behavior can be shaped to peck at a button to guide a rocket, profs’ behavior can be shaped to ‘likes.’
The Skinnerian operant
Reply-all messages count as a publication.
Come to think of it, my largest readership is for my reply-all messages. ☹️
Is this a campus-wide issue? I thought it was only like this in Scarfe.
Have they assigned the control of room temperature to WorkDay?
Is it available in Canadian?
Those behaviour / behavior and doughnut / donut trip me up every time. Although I suspect there may not be much concern for doughnuts in your book, behaviour may come up once or twice. 🤭
All silliness aside, congratulations.
For example, throughout the 1990s Don Zimmerman and I extended his ‘75 paper and formulated a classical test theory as an abstract mathematical model, using concepts in measure theory, probability theory, and functional analysis. The central part of that program appeared in 2001.
[2/3]