Sign in

Yongnam Kim

@ykims.bsky.social
347 followers 139 following 26 posts

Education researcher interested in causal inference & DAGs | Seoul National University

PostsRepliesMedia
Reposted by Yongnam Kim
Michael Friendly @datavisfriendly.bsky.social · 13/09/2026
#TodayinHistory #dataviz #OTD 📊 📅Sep 10, 1885 Francis Galton introduced the idea of regression (toward the mean) (In his presidential lecture to the Anthropology Section of the British Association, perhaps the most influential such in statistical history)
Galton's figure plotting the height of children vs. the average height of their parents, titled "Rate of Regression in Hereditary stature".
Two lines are labeled: Mid-parents and Children. Labels in the plot say: When Mid-Parents are taller than mediocraty, their children tend to be shorter than they.
Two photos of Francis Galton, left: in profile; right: full face, as if posing for mug shots.. These were related to Galtn's studies of criminality, now standard in police procedure.
1162
Reposted by Yongnam Kim
Paul Hünermund @p-hunermund.com · 29/06/2026
Why is that bad though?
6283
Yongnam Kim @ykims.bsky.social · 09/06/2026
Once I accept the structure A_M (or A_Y) → A, it feels strange to me that A’s effect gets separated into the component effects of A_M and A_Y, when A is the common effect of them, not their parent.
010
Yongnam Kim @ykims.bsky.social · 09/06/2026
The examples here suggest to me the reverse direction, A_M (or A_Y) -> A. Attending meetings and changing diet together constitute participating in the weight-loss program, rather than the program causing them. The causal relationship between A and A_M (or A_Y) always confuses me. Not you?
120
Reposted by Yongnam Kim
Protzko @protzko.bsky.social · 20/05/2026
Teaching DAGs sounds easy, until students actually have to make one themselves. A new tutorial lays out simple steps to make it clear and easy for students to understand. Draw process, selection, outcome, confounds. From Nathan J. Quimpo & Peter M. Steiner files.osf.io/v1/resources... #psych
14516
Yongnam Kim @ykims.bsky.social · 14/03/2026
maybe, that's why Pearl classifies PS as an associational concept, not a causal one.
010
Yongnam Kim @ykims.bsky.social · 14/03/2026
Another view on propensity score (PS) in DAGs. academic.oup.com/ije/article-... PS is a mediator like covariates -> PS -> treatment. I think we are obsessed with seeing how conditioning on PS makes the balancing property in DAGs. Maybe d-separation is simply not the right tool. Just accept it.
academic.oup.com
Making DAGs even more useful: using augmented causal diagrams to depict counterfactual, study design, measurement, analytical, and interventional features
Since their mainstream introduction in the 1990s, causal diagrams, including directed acyclic graphs (DAGs), have been increasingly used to depict our caus
191
Yongnam Kim @ykims.bsky.social · 05/03/2026
"No Manipulation, No Causation"? She's crying because of her height. How to intervene on height doesn't matter. Height clearly causes her crying. Maybe studying it isn't useful, but utility is not ontology.
041
Reposted by Yongnam Kim
koenfucius @koenfucius.bsky.social · 27/02/2026
You may very well think that Philosophy is not a fact-based discipline. Bryan Frances will disabuse you of such thoughts, offering 200 philosophical facts in evidence: buff.ly/lDX367Y
051
Yongnam Kim @ykims.bsky.social · 25/02/2026
also, here is a DAG for the classical DiD by Card & Krueger (1994) osf.io/preprints/ps... Presenting it to an Econ audience, one question was, this seems not DiD bc there is no time and the interaction. Another fun claim of it: DiD can be, indeed has been used to deal with the POSITIVITY violation.
osf.io
OSF
120
Yongnam Kim @ykims.bsky.social · 25/02/2026
Lots to say on this, but one thing recently came to my mind is the isomorphism between “wide format” & “long format” for panel datasets. The latter treats time as a variable, which confuses the issue, at least conceptually. DiD with wide format is clear and here: journals.sagepub.com/doi/10.1177/...
journals.sagepub.com
Gain Scores Revisited: A Graphical Models Perspective - Yongnam Kim, Peter M. Steiner, 2021
For misguided reasons, social scientists have long been reluctant to use gain scores for estimating causal effects. This article develops graphical models and g...
155
Yongnam Kim @ykims.bsky.social · 20/02/2026
that's what people think about true PS. For estimated or computed PS, different representations have been considered. One is Vs -> PS <- X, the other is just Vs -> PS. I used to think the former was right, but now I think it's the latter.
010
Yongnam Kim @ykims.bsky.social · 20/02/2026
But I certainly agree that DAGs have weaknesses. Not everything can be explained with DAGs.
000
Yongnam Kim @ykims.bsky.social · 20/02/2026
the coefs carry some "dependence" on A. But once the coefs are determined, PS is computed purely as a function of covariates by the coefs. DAG encodes that functional relationship, not the process by which the coef were determined. A's role in estimation doesn't justify A -> PS.
110
Yongnam Kim @ykims.bsky.social · 19/02/2026
and it contains probably the 1st DAG representation of 2SLS, differing from Wald estimation. The 2SLS analogy was key to convincing me about the PS case. In the DAG, Cov(A-hat, Y)/Var(A-hat) = tau follows cleanly from path-tracing rules, and adding A → A-hat would break the IV identification.
230
Yongnam Kim @ykims.bsky.social · 19/02/2026
How to draw propensity scores (PS) in DAGs? Some (me also) claim it is like "treatment -> PS <- covariates", since in order to compute PS we need both treatment and covariates. This view has confused me for so long, and now I think I was wrong. My letter here: track.smtpsendmail.com/9032119/c?p=...
academic.oup.com
Reconsidering the graphical representation of propensity scores in causal diagrams
I read with interest Mansournia et al.’s article ‘Balancing scores and causal diagrams’ [1]. While their effort to use directed acyclic graphs (DAGs) to il
41711
Yongnam Kim @ykims.bsky.social · 26/12/2025
A bit late, but you might find this interesting, osf.io/preprints/ps.... I think we have the same graph about Lord’s paradox.
osf.io
OSF
110
Yongnam Kim @ykims.bsky.social · 22/10/2025
This leads to an embarrassing thought: what I draw in my DAGs might itself be the result of a collider in some meta-DAG of the universe. I drew Sex → Weight and was so sure of the structure. But in a higher-order universe, this might itself be the result of collider conditioning.
020
Yongnam Kim @ykims.bsky.social · 22/10/2025
What does “unconditional” really mean? P(data) seems unconditional, and P(data | boys) conditional. But imagine an alien landing on Earth and seeing P(data). It says, “Oh, so you’re conditioning on humans, not tigers.” Every “unconditional” is just conditional on a world we take for granted.
130
Yongnam Kim @ykims.bsky.social · 16/05/2025
We’re too obsessed with decomposing direct and indirect effects in mediation. "mediation should not be understood in terms of decomposition...Once the priority of research questions is established, the practical irrelevance of statistical effect decomposition directly follows" osf.io/preprints/ps...
osf.io
OSF
050
Yongnam Kim @ykims.bsky.social · 02/05/2025
a fun part is, these two approaches might give conflicting results about the effect of T. I think this can be another version of Lord's paradox.
000
Yongnam Kim @ykims.bsky.social · 02/05/2025
I think your approach is ok. You just defined your question as the effect of T on Y/X, and there’s nothing wrong with it. But it might be good to think about why you're using Y/X. If you want to account for the role of X, another option is Y~T+X, which gives the effect of T on Y holding X constant.
100
Yongnam Kim @ykims.bsky.social · 01/05/2025
Card & Krueger's (1994) minimum wage study may be such an extreme case of confounding: "State" (NJ vs. PA), a confounder, perfectly correlates with the causal variable "minimum wage." Their interest was in the effect of minimum wage on employment, not the effect of restaurants' state location.
0152
Reposted by Yongnam Kim
Riccardo Fusaroli @fusaroli.eurosky.social · 07/02/2025
Looking for a tool to more easily draw your DAGs and reason on them? Try PV-dagger (pvverse.github.io/pv_dagger/). Specifically designed by @fusarolimichele.bsky.social to deal with the complex DAGs involved in pharmacovigilance, helps positioning and color-coding confounds, measurement errors, etc
pvverse.github.io
Visualization of Causal Structures in Pharmacovigilance Data Using DAGs
The PVdagger package provides tools for creating and visualizing Directed Acyclic Graphs (DAGs) with various biases and paths. This package is particularly useful for researchers and signal managers i...
092
Yongnam Kim @ykims.bsky.social · 27/01/2025
A key insight is the equivalence btw suppressors and instrumental variables. Yes, DAGs are useful for understanding why S is zero-related with Y, yet can increase the overall prediction.
052
Reposted by Yongnam Kim
Quantitude the Podcast @quantitude.bsky.social · 10/01/2025
18113
Yongnam Kim @ykims.bsky.social · 25/01/2025
Card & Krueger’s minimum wage study may be a real example of a positivity violation. Their DiD addresses positivity, not unconfoundedness. osf.io/preprints/ps...
osf.io
OSF
120
Reposted by Yongnam Kim
Richard McElreath 🐈‍⬛ @rmcelreath.bsky.social · 21/01/2025
This sounds like the same error I blogged about a few years ago, the common error of trying to control for population (or body size or many etc) by dividing the outcome variable by it. Props to the authors for seeking review and taking the issue seriously. Role models for us all.
elevanth.org
There Are No Magic Outcome Variables
I'm very sorry but I am going to write about statistics and causal inference again. I'd much rather be doing science. I'll make it brief. I was reading a preprint about the relationship between kinsh...
510120
Yongnam Kim @ykims.bsky.social · 19/01/2025
Why HIGHER? If not, A² also be part of the Y model, implying A² → Y, which violates the exclusion restriction. This shows why the DAG representation suggested in shorturl.at/Tj8am is useful. A² = A × A can be described in DAGs, offering intuition for analysis mechanics.
shorturl.at
British Journal of Mathematical and Statistical Psychology | Wiley Online Library
Interaction analysis using linear regression is widely employed in psychology and related fields, yet it often induces confusion among applied researchers and students. This paper aims to address thi....
010
Yongnam Kim @ykims.bsky.social · 19/01/2025
Clear from the DAG, A² acts as an instrumental variable (conditional on A), enabling the identification of the M → Y effect even with U. This is what shorturl.at/1TgCm showed: mediation analysis can be valid (even with U) if the M model has a higher order of A than the Y model.
120
Yongnam Kim @ykims.bsky.social · 02/12/2024
Very happy to share this final version with you. Thank you! ;-)
010
Yongnam Kim @ykims.bsky.social · 02/12/2024
Easy to see why the cor btw the first-order and interaction terms (indicating collinearity) after centering becomes zero (though this is not the reason for centering); why centering X1 only (not X2) change the coef on X2​ while leaving the coefs on X1 and the (centered) interaction term unchanged.
040
Yongnam Kim @ykims.bsky.social · 02/12/2024
DAGs (causal graphs) can be used to understand the mechanics of linear interaction analysis. See more here: bpspsychub.onlinelibrary.wiley.com/doi/10.1111/...
bpspsychub.onlinelibrary.wiley.com
British Journal of Mathematical and Statistical Psychology | Wiley Online Library
Interaction analysis using linear regression is widely employed in psychology and related fields, yet it often induces confusion among applied researchers and students. This paper aims to address thi...
1121