# Causal Formalism and RCTs

**URL:** <https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854>\
**Category:** causal inference\
**Tags:** interpretation, design\
**Created:** [September 10, 2026, 2:33pm UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854 "2026-09-10T14:33:46Z")\
**Posts on this page:** 7\
**Page:** 2

<div class="post-metadata">

**Author:** ![f2harrell](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/f2harrell/32/165_2.png) [@f2harrell](https://discourse.datamethods.org/u/f2harrell)\
**Post date:** [September 14, 2026, 2:30pm UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/21 "2026-09-14T14:30:52Z")

</div>

You’ve made your point, let’s move on to other considerations. Thanks.

---

<div class="post-metadata">

**Author:** ![llynn](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/llynn/32/2827_2.png) [@llynn](https://discourse.datamethods.org/u/llynn)\
**Post date:** [September 14, 2026, 2:57pm UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/22 "2026-09-14T14:57:47Z")

</div>

Indeed I have!

I know many statisticians wish I was silent but you chose the very broad title “causal formalization and RCTs”. If you wanted to keep it limited to the “safe stuff” a different title was probably in order.

Very well, I’ll exit the discussion.

---

<div class="post-metadata">

**Author:** ![f2harrell](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/f2harrell/32/165_2.png) [@f2harrell](https://discourse.datamethods.org/u/f2harrell)\
**Post date:** [September 14, 2026, 3:38pm UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/23 "2026-09-14T15:38:30Z")

</div>

No need to exit the discussion, and this need not stay with the “safe stuff”. I just want to limit repetition. You could factor out a good deal of what you write into orthogonal nuggets 🙂

---

<div class="post-metadata">

**Author:** ![samw235711](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/samw235711/32/4255_2.png) [@samw235711](https://discourse.datamethods.org/u/samw235711)\
**Post date:** [September 27, 2026, 11:31pm UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/24 "2026-09-27T23:31:58Z")

</div>

I hoped to add: although the no measure confounders assumption is formidable, often it’s a trade-off between making it and not doing an analysis at all. Sometimes better to do the analysis, especially when one can’t randomize (eg, tobacco use).

---

<div class="post-metadata">

**Author:** ![Pavlos\_Msaouel](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/pavlos_msaouel/32/739_2.png) [@Pavlos\_Msaouel](https://discourse.datamethods.org/u/Pavlos_Msaouel)\
**Post date:** [September 28, 2026, 1:18am UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/25 "2026-09-28T01:18:16Z")

</div>

> [@f2harrell](#):
>
> As [AP Dawid has elegantly shown](https://www.jstor.org/stable/2669377), a cohesive analysis that involves “what might have happened had a patient received the treatment they didn’t receive” requires one to know or well-estimate the variance of Y(1)-Y(0). The joint distribution between a patient’s two potential outcomes can never be estimated or checked, even in principle, and even for infinite sample sizes.

Your Dawid reference and advantage of relying on observables are far more pertinent than any of us interested in causal inference realized in the pre-LLM era. Folks should pay close attention – there is magic to unfold.

---

<div class="post-metadata">

**Author:** ![llynn](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/llynn/32/2827_2.png) [@llynn](https://discourse.datamethods.org/u/llynn)\
**Post date:** [September 30, 2026, 2:42am UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/26 "2026-09-30T02:42:10Z")

</div>

> [@f2harrell](#):
>
> this need not stay with the “safe stuff”

Very well.

Recognition that cause-agnostic RCTs (CARs) can be an unsafe standard methodology makes causal formalism necessary. But why are they unsafe? They are based on poor priors. If the eligibility rule is only weakly coupled to the mechanism the treatment is supposed to act on, the prior probability that a positive result is a _true disease-level effect_ is lower.

Causal formalism distinguishes who enters an RCT from the disease process the treatment actually affects. Complete mechanistic knowledge is not required, but the entry criteria must be linked, at least probabilistically, to the process being treated. This is consistent with a Bayesian strengthening of RCT design: when the treatment and eligibility gate are poorly matched, the probability of a true treatment effect is lower, increasing the probability that a positive result is false. It also increases the risk of “false transport” (applying a false positive RCT result to clinical Guidelines.

The “reproducibility crisis” is a misnomer. The proper term is a “false transportability” crisis. This is cause by standardized cause agnostic gates which render insufficient priors .Bayesian approaches which do not interrogate the gate (like the failed 2025 RE MAP CAP) are “Bayesian facades”.

---

<div class="post-metadata">

**Author:** ![f2harrell](https://discourse.datamethods.org/user_avatar/discourse.datamethods.org/f2harrell/32/165_2.png) [@f2harrell](https://discourse.datamethods.org/u/f2harrell)\
**Post date:** [September 30, 2026, 11:41am UTC](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854/27 "2026-09-30T11:41:50Z")

</div>

Getting back to the generic topic related to Pearl’s causal calculus (primarily rung 3 in the causal ladder) I am getting a more solid sense that many researchers are using this formalism so that they can feel good about stating assumptions out in the open, never minding how unlikely those assumptions are to hold. Sometimes the formalisms used also change the original clinical question to something less relevant. I am working now on a blog article that delves into this in detail. The article will contrast falsifiable but untestable assumptions with fully data-informed analysis, and unobservables with observables, in the context of truncation by death.

[Previous page](https://discourse.datamethods.org/t/causal-formalism-and-rcts/28854.md?page=1)