Julia M. Rohrer: If an effect falls in a forest and no one is there to determine the mechanism, is it even causal? New post in which I try to clarify some things--claims about causal effects are indifferent to mechanisms; heterogeneity does not invalidate average estimates. www.the100.ci/2024/06/26/s...

If an effect falls in a forest and no one is there to determine the mechanism, is it even causal? New post in which I try to clarify some things--claims about causal effects are indifferent to mechanisms; heterogeneity does not invalidate average estimates. www.the100.ci/2024/06/26/s...

Sometimes a causal effect is just a causal effect (regardless of how it’s mediated or moderated)www.the100.ci TL;DR: Tell your students about the potential outcomes framework. It will have (heterogeneous) causal effects on their understanding of causality (mediated through unknown pathways), I promise. It’...

I'm gonna reuse this 🤣

Originally I had only the goose meme; @malte.the100.ci suggested to add an interaction plot although I think he wanted it to be the plot itself running away, wearing shopped shoes and all.

I've encountered papers averaging reaction time of two totally different tasks. 🤦

Works for conjoint analysis estimands too, but you need two geese.

oh no

Another 2 cents on looking to phantom HTE to save a "null" finding: statsepi.substack.com/p/perils-of-...

Perils of Personalized Treatment Effectsstatsepi.substack.com Episode #456: Hidden Subgroups

You could refer your children to this blog post with their never-ending why-questions

oof ouch owie my head

Another banger Julia! Interesting point on 'starts from the opposite notion that causal effects are the same for everyone'. Seems similar in experimental design literature. Often 'assuming this model [where all individual causal effects are identical], then the optimal design…'

In psych it seems a bit split; like clearly some interpretations imply that people take homogeneity as the default, but then people are also very quick to point out that things are very complicated, everybody is an individual etc

Oh me pick me I know this one! There’s a constant union betwixt the cause – UNIMPRESSIVE ATE DESPITE STRONGLY HELD HUNCHES – and effect – I THINK I’LL FIND IT’S A NOT MORE COMPLICATED THAN ATE CAPTURES

I think mechanisms are the best way we can gauge whether a model producing a causal effect estimate are in accordance with theory (i.e., credible)

Reminds me of Wittgenstein for some reason: "a wheel that can be turned though nothing else moves with it is not part of the mechanism”(§271)."

This is a good introduction to the potential outcomes framework, though I worry that the potential outcomes framework fails to zoom in on some of the regularities that make causal inference more tractable.

Perhaps most importantly, reality tends to have high skewness/high kurtosis/long tails and sparsity, but also high dimensionality. These features make some methods more feasible and others less feasible.

Post