the people, they won’t stop talking about it, they’re demanding Posttreatment Bias Magazine
There should be a whole magazine about policymakers not understanding controls.
May 1, 2026 at 7:39 PM
the people, they won’t stop talking about it, they’re demanding Posttreatment Bias Magazine
Me at work: you shouldn’t condition on posttreatment variables as it biases inferences, see Montgomery et al. (2018).
Me at home: look at how sweet the kids are with each other when they’re not fighting!
Me at home: look at how sweet the kids are with each other when they’re not fighting!
August 2, 2025 at 12:38 PM
Me at work: you shouldn’t condition on posttreatment variables as it biases inferences, see Montgomery et al. (2018).
Me at home: look at how sweet the kids are with each other when they’re not fighting!
Me at home: look at how sweet the kids are with each other when they’re not fighting!
Always happy to help 😘 here is the canonical paper on posttreatment bias, highly recommended: onlinelibrary.wiley.com/doi/full/10....
October 17, 2023 at 10:18 AM
Always happy to help 😘 here is the canonical paper on posttreatment bias, highly recommended: onlinelibrary.wiley.com/doi/full/10....
In quantitative social science analyses, it’s striking how often there’s uncertainty on whether various correlates of outcomes are CONFOUNDERS or MEDIATORS.
In past ~20 years, we finally realized how pervasive & pernicious posttreatment control is. Collective learning on that is good.
But…
In past ~20 years, we finally realized how pervasive & pernicious posttreatment control is. Collective learning on that is good.
But…
June 7, 2025 at 4:30 PM
In quantitative social science analyses, it’s striking how often there’s uncertainty on whether various correlates of outcomes are CONFOUNDERS or MEDIATORS.
In past ~20 years, we finally realized how pervasive & pernicious posttreatment control is. Collective learning on that is good.
But…
In past ~20 years, we finally realized how pervasive & pernicious posttreatment control is. Collective learning on that is good.
But…
A style of paper: In isolation, it's really just an incident of somewhat confused statistical analyses (e.g., people being confused about the precise estimand, controling for posttreatment variables etc.). But then these are repeated across a large number of data-sets with a huge author team.>
June 23, 2025 at 8:25 AM
A style of paper: In isolation, it's really just an incident of somewhat confused statistical analyses (e.g., people being confused about the precise estimand, controling for posttreatment variables etc.). But then these are repeated across a large number of data-sets with a huge author team.>
Causal inference iceberg!
What's missing?
What's missing?
February 26, 2025 at 1:40 PM
Causal inference iceberg!
What's missing?
What's missing?
One test I’ve been considering is the most old fashioned but tried and true: show the results both ways and reflect on implications of results. At very least, we can’t just declare posttreatment control or confounding on assumptions alone.
June 7, 2025 at 4:36 PM
One test I’ve been considering is the most old fashioned but tried and true: show the results both ways and reflect on implications of results. At very least, we can’t just declare posttreatment control or confounding on assumptions alone.
Do you think that learning more about causal inference is not worth it because you're running experiments anyway, or because you're interested in predictive questions? In that case, I've written a paper just for you, out now in SPPC: compass.onlinelibrary.wiley.com/doi/10.1111/...
March 2, 2024 at 5:25 AM
Do you think that learning more about causal inference is not worth it because you're running experiments anyway, or because you're interested in predictive questions? In that case, I've written a paper just for you, out now in SPPC: compass.onlinelibrary.wiley.com/doi/10.1111/...
Easy solution, include an attention check and exclude based on that, thus trading data quality for posttreatment bias 😙
September 29, 2025 at 5:44 AM
Easy solution, include an attention check and exclude based on that, thus trading data quality for posttreatment bias 😙
1. Má už 3,7 kilo a zvýšila se dávka antivirotik, takže teď dávám 1 + 0,5 + 0,25 ta lety denně, což je sice nepraktické, ale mám už grif. Dostala ostropestřec na ochranu jater.
2. Sílí, roste a šílí. :D ještě budeme podávat léky necelých 50 dní a pak ještě posttreatment.
2. Sílí, roste a šílí. :D ještě budeme podávat léky necelých 50 dní a pak ještě posttreatment.
No a vůbec co novýho s marodem? Chceme aktuální info 😺
March 8, 2025 at 3:25 PM
1. Má už 3,7 kilo a zvýšila se dávka antivirotik, takže teď dávám 1 + 0,5 + 0,25 ta lety denně, což je sice nepraktické, ale mám už grif. Dostala ostropestřec na ochranu jater.
2. Sílí, roste a šílí. :D ještě budeme podávat léky necelých 50 dní a pak ještě posttreatment.
2. Sílí, roste a šílí. :D ještě budeme podávat léky necelých 50 dní a pak ještě posttreatment.
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
doi.org/10.1016/j.cc...
doi.org/10.1016/j.cc...
May 21, 2025 at 1:50 AM
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
Maybe you’re thinking of posttreatment covariates (variables assessed after treatment)? That can induce something called posttreatment bias, which I discuss here: compass.onlinelibrary.wiley.com/doi/10.1111/...
Causal inference for psychologists who think that causal inference is not for them
Correlation does not imply causation and psychologists' causal inference training often focuses on the conclusion that therefore experiments are needed—without much consideration for the causal infer....
compass.onlinelibrary.wiley.com
May 13, 2024 at 4:42 PM
Maybe you’re thinking of posttreatment covariates (variables assessed after treatment)? That can induce something called posttreatment bias, which I discuss here: compass.onlinelibrary.wiley.com/doi/10.1111/...
Teaching Masters-level statistics for umpteenth time:
Why do econometrics texts spend lots of space on collinearity when it’s almost never a problem & tests are easy?
Yet econometrics texts spend near zero space on posttreatment control even though it’s a pervasive problem & it’s hard to test?
Why do econometrics texts spend lots of space on collinearity when it’s almost never a problem & tests are easy?
Yet econometrics texts spend near zero space on posttreatment control even though it’s a pervasive problem & it’s hard to test?
February 7, 2025 at 5:10 PM
Teaching Masters-level statistics for umpteenth time:
Why do econometrics texts spend lots of space on collinearity when it’s almost never a problem & tests are easy?
Yet econometrics texts spend near zero space on posttreatment control even though it’s a pervasive problem & it’s hard to test?
Why do econometrics texts spend lots of space on collinearity when it’s almost never a problem & tests are easy?
Yet econometrics texts spend near zero space on posttreatment control even though it’s a pervasive problem & it’s hard to test?
Friends don't let friends ruin their experiments by dropping participants who fail manipulation checks - new illustration in JEPS https://tinyurl.com/3zxt6u6w
See our AJPS - How Conditioning on Posttreatment Variables Can Ruin Your Experiment and What to Do about It
https://tinyurl.com/yz4dbbuw
See our AJPS - How Conditioning on Posttreatment Variables Can Ruin Your Experiment and What to Do about It
https://tinyurl.com/yz4dbbuw
July 5, 2023 at 12:31 PM
Friends don't let friends ruin their experiments by dropping participants who fail manipulation checks - new illustration in JEPS https://tinyurl.com/3zxt6u6w
See our AJPS - How Conditioning on Posttreatment Variables Can Ruin Your Experiment and What to Do about It
https://tinyurl.com/yz4dbbuw
See our AJPS - How Conditioning on Posttreatment Variables Can Ruin Your Experiment and What to Do about It
https://tinyurl.com/yz4dbbuw
And while I’m still mostly on side of worrying more about posttreatment control in obsv data & analyses, I wonder if we could all agree on a more standardized view of discerning between these concerns.
June 7, 2025 at 4:34 PM
And while I’m still mostly on side of worrying more about posttreatment control in obsv data & analyses, I wonder if we could all agree on a more standardized view of discerning between these concerns.
I worry a bit the pendulum has swung such that people are excusing themselves from adjusting for plausible confounders by declaring the threat of posttreatment control bias.
Then they show some surprising or contrarian or flashy finding that only holds up if you omit what we used to control for.
Then they show some surprising or contrarian or flashy finding that only holds up if you omit what we used to control for.
June 7, 2025 at 4:32 PM
I worry a bit the pendulum has swung such that people are excusing themselves from adjusting for plausible confounders by declaring the threat of posttreatment control bias.
Then they show some surprising or contrarian or flashy finding that only holds up if you omit what we used to control for.
Then they show some surprising or contrarian or flashy finding that only holds up if you omit what we used to control for.
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
doi.org/10.1016/j.cc...
doi.org/10.1016/j.cc...
May 21, 2025 at 2:15 AM
Self-assembly of disassemblable supramolecular organic frameworks for doxorubicin delivery, photofrin posttreatment phototoxicity inhibition and heparin neutralization
But the test I really think we need - and aren’t seeing enough - is if you want to justify omitting a control bc of posttreatment bias, you should show EVIDENCE of mediation. If you omit Z because X -> Z -> Y, then prove X-> Z and Z -> Y.
Is anyone else wondering about this these days?
Is anyone else wondering about this these days?
June 7, 2025 at 4:38 PM
But the test I really think we need - and aren’t seeing enough - is if you want to justify omitting a control bc of posttreatment bias, you should show EVIDENCE of mediation. If you omit Z because X -> Z -> Y, then prove X-> Z and Z -> Y.
Is anyone else wondering about this these days?
Is anyone else wondering about this these days?