Yongnam Kim

Associate Professor, Dept. of Education, Seoul National University | Researcher in Causal Inference

[Yongnam Kim @y2silence]

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 https://t.co/A3919fIxfB is useful. A² = A × A can be described in DAGs, offering intuition for analysis mechanics.

[Yongnam Kim @y2silence]

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 https://t.co/svoi5djzxH: mediation analysis can be valid (even with U) if the M model has a higher order of A than the Y model.

[Yongnam Kim @y2silence]

Why does centering X1 only (not X2) change the coef on X2​ while leaving the coefs on X1 and the (centered) interaction term unchanged? For those who enjoy DAGs, here is a slightly modified visual explanation we provided (https://t.co/M1w86PSsqd)

[Matt Blackwell @matt_blackwell]

A fun fact about regression that many know but maybe is new to you:
If you have an interaction bw continuous X1 and binary X2, mean-centering X1 will make the coefficient on X2 be its marginal effect when X1 is at its mean level rather than 0 without changing the interaction

[Yongnam Kim @y2silence]

DAGs can be used to understand the mechanics of linear interaction analysis. Easy to see the zero correlation between the first-order and interaction terms after centering (though this is not the reason for centering). See more here: https://t.co/zfRnw50PRL

[Soo Jung (Hellena) La @SJHellena]

I saw earlier posts on DiD, so I thought some of you might be interested in my recent work in causal inference!

Our paper illustrates how DiD can be used, albeit positivity violation: https://t.co/AWYPHOzabC

Go check it out if interested!

#DiD #causalInference #lordsParadox

[Yongnam Kim @y2silence]

Typo correction. <Can two variables have the same values but different impacts? Revisiting Card & Krueger (1994) and Lord’s (1967) paradox> A new perspective & a new answer on Lord’s paradox

[PsyArXiv bot v2 @PsyArXiv_bot_v2]

Can Two Variables Have the Same Values but Different Impacts? Revising Card & Krueger (1994) and Lord's (1967) Paradox https://t.co/GJnkYhqIEB

[Yongnam Kim @y2silence]

These definitions may help to resolve some confusion b/t #DIF (association) and item bias (causation) in the psych measurement literature. Interpretation of spurious differential item functioning: https://t.co/Tj31XARCUk

[Yongnam Kim @y2silence]

@KristopherJPre1 thanks for helping me out. Just wanted to understand the new concept from the paper. Seems like there is something more behind it, I didn't realize at first glance. Thanks!

[Yongnam Kim @y2silence]

Preacher’s (2006) e.g. to explain “fitting propensity”: A model’s property to fit data better than other. With randomly generated data, he showed that B fits the data better than A.

Do we really need the concept, though? Look at the v-str in A, implying zero-cor b/w V1 & V2.

[Yongnam Kim @y2silence]

@oscar_olvera100 Thanks! I think I will also like it!

[Yongnam Kim @y2silence]

VanderWeele’s critique of psychological research practices based on factor analysis. https://t.co/NaAE1pAoge It seems to be related to the debate on bifactor models.

[Yongnam Kim @y2silence]

@JohannesTextor Hmm.. true. Then maybe fitting propensity might capture something different.. or I think the two constraints may result in different consequences for the fit indexes.

[Yongnam Kim @y2silence]

@JohannesTextor https://t.co/5GtoObbRqp

[Yongnam Kim @y2silence]

@JohannesTextor Thanks, good point. I think the global fit index he used just compared the covariance matrices. Not considered conditional (in)dependences.

[Yongnam Kim @y2silence]

@oscar_olvera100 Can u recommend one such paper from health researchers? I am wondering how clearly they discussed the causal meaning of SEM.

[Yongnam Kim @y2silence]

@oscar_olvera100 Thanks for the comment. I read them as a general critique of factor analysis practices, not just about unidimensionality. The point I enjoyed was the causal meanings of factor analysis results. Also, well demonstrated here: https://t.co/1PyS9dvKQV by @mijkenijk

[Judea Pearl @yudapearl]

"Race, COVID Mortality, and Simpson's Paradox." Dana Mackenzie, my co-author in #Bookofwhy, has discovered another case of Simpson's Paradox in COVID data, and has posted an interesting analysis of its implications here:
https://t.co/3MfnyC17x5

[James E. Pustejovsky @jepusto]

What in Hell is this? I'm dumbfounded by how a policy could be both so malicious and so idiotic.