gregoryfaletto.com
Greg Faletto
@gregoryfaletto.com
370 followers 700 following 510 posts
Statistician, Data Scientist.
Posts Media Videos Starter Packs
Pinned
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · Jul 5
Interested in doing some causal inference over the long weekend? Version 1.5.0 of the {fetwfe} package for difference-in-differences under staggered adoptions is now on CRAN!

Details on what's new below:

cran.r-project.org/web/packages...
fetwfe: Fused Extended Two-Way Fixed Effects
Calculates the fused extended two-way fixed effects (FETWFE) estimator for unbiased and efficient estimation of difference-in-differences in panel data with staggered treatment adoption. This estimato...
cran.r-project.org
1 1 11
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 27m
"Snake eating its own tail" deals like Nvidia investing in OpenAI certainly seem to suggest "more money floating around than can be put to productive use."
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 10h
I think it's fair to say later, I think the 2008 election was basically unaffected by any of the dynamics that we today attribute to social media/online misinformation/unabashed lying etc.

(I was tempted to say the same for 2012 but birtherism was before that.)
1 1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 11h
Technically a sufficiently motivated Darren can
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 2d
I don't even get the logic. Why can't ICE take care of themselves? Too much time spent on rappelling training, not enough self-defense? Who's coming next to protect the poor widdle Texas National Guard from antifa?
4
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
"Is Method A better than B *in this case*" depends on the available data and the estimand.

But of course there are many RCTs that are not wastes of money--they are more credible causal estimates than the best possible observational study.
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
Seeing some of your other posts, I agree that external validity often is ignored when we think about method assumptions, and that observational studies that we care about in practice are often ill-suited for an externally valid experiment.
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
Sure, but in a pop culture sense, I can get why after going from basically no free public text- or image-generating models to having GPT 3.5 and several latent diffusion models within the span of a year, the general public becomes receptive to the idea "AI is finally here"
1 5
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
Well it's good for them that they've been carefully currying favor with military leaders. Nobody in the administration just wasted literally hundreds of generals' and admirals' time to tell them a bunch of stupid bullshit this week by any chance, did they?
1 19
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
Understanding the relationship between the population you can actually sample and the population you're actually interested in is definitely important.
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 3d
Cool, thanks for sharing!
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
It seems perfectly reasonable to not believe in counterfactual outcomes for a single individual. It's just that if you don't believe in that, then all causal inference (including RCTs) is a fiction and there's nothing to do. Most people don't buy that so we have to swallow the existence of ITEs
1 5
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
* It's reasonable to say these estimators are in a definitive hierarchy of credibility.
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
* There are estimators that work under either exogenous randomization or other assumptions like unconfoundedness and others (e.g. regression adjustment).
* In some cases it's possible to nest the assumptions required for a pair of estimators.
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
I'm curious where you'd stop agreeing:

* Exogenous randomization is useful for identifying causal effects.
* There are estimators that identify causal effects only under exogenous randomization (e.g. difference of means).
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
Sure but the same problems can in principle exist with observational data (that we can observe).
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
Do you mean the ATE vs the ATT? With uncofoundedness we can identify both, yes?
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
Do you want to elaborate? If Method 1 requires SUTVA and something else too and Method 2 only requires SUTVA, in any traditional sense of the word that's a "hierarchy."

Here's another example: unconfoundedness implies conditional parallel trends but not the other way around.
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
I mean, some assumptions are weaker than others. If Assumption A implies Assumption B but not the other way around, that's a hierarchy.
2 1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 4d
He's not running for re-election
2
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 6d
Am I going crazy or has Paramount bought The Free Press and made Bari Weiss the head of CBS News like every day for the last two weeks?
1 1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 6d
One thing that's tricky is that even if our only concern is time-invariant covariates, it's unclear why differencing ought to be the right operation to "undo" it--that's kind of smuggling in an additive functional form assumption. How do we know the right operation isn't dividing, or anything else?
1
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 7d
But the substance of his economic policies, while not perfect, was the best of any president in generations, especially when you grade on the curve of the razor-thin margins he had.
1 6
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 7d
You're imagining somebody different in your head. Personally I think Biden's presidency was overall a failure because he failed at the most important crisis of his presidency, which was stamping out MAGA. And he clearly miscalculated on the politics of his economic policies--that's on him.
1 9
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 7d
Your argument is with the historical record not us bsky.app/profile/greg...
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 7d
Yeah it really sucked when Biden didn't get his way on the child tax credit. en.wikipedia.org/wiki/Build_B...
Screenshot of a Wikipedia article section discussing Senator Joe Manchin’s opposition to extending the child tax credit. The text explains that Senate Democrats supported the enhanced credit, citing reductions in child poverty and food insecurity. Manchin demanded stricter work requirements and criticized reliance on temporary funding. He rejected claims that he wanted to eliminate the credit, calling them false, but expressed concerns about misuse of funds. The Census Bureau is cited as reporting that most families used the payments for essentials like food, clothing, utilities, and education. Numerous reference citations in blue are interspersed throughout the text.
1 10
gregoryfaletto.com
Greg Faletto @gregoryfaletto.com · 7d
Yeah it really sucked when Biden didn't get his way on the child tax credit. en.wikipedia.org/wiki/Build_B...
Screenshot of a Wikipedia article section discussing Senator Joe Manchin’s opposition to extending the child tax credit. The text explains that Senate Democrats supported the enhanced credit, citing reductions in child poverty and food insecurity. Manchin demanded stricter work requirements and criticized reliance on temporary funding. He rejected claims that he wanted to eliminate the credit, calling them false, but expressed concerns about misuse of funds. The Census Bureau is cited as reporting that most families used the payments for essentials like food, clothing, utilities, and education. Numerous reference citations in blue are interspersed throughout the text.
1 16