Dan de Kadt
@dandekadt.bsky.social
Social and data science at the London School of Economics
Democracy, behaviour, meta-science, 🇿🇦🇺🇲
Won't interact with anon accounts.
www.ddekadt.com
Democracy, behaviour, meta-science, 🇿🇦🇺🇲
Won't interact with anon accounts.
www.ddekadt.com
I think this is one perspective (not a bad one) but not the only. The alternative view asks "what do we need to believe about the data generating process (or assignment process) to admit a causal interpretation to the parameters we estimate?"
November 11, 2025 at 2:17 PM
I think this is one perspective (not a bad one) but not the only. The alternative view asks "what do we need to believe about the data generating process (or assignment process) to admit a causal interpretation to the parameters we estimate?"
Yes another good suggestion!
November 11, 2025 at 11:37 AM
Yes another good suggestion!
I think the general lessons are:
1. Be clear about your goal (_estimand_)
2. Be clear about your analytical method (_estimator_)
3. Be clear about the assumptions that connect method to goal (_design_)
1. Be clear about your goal (_estimand_)
2. Be clear about your analytical method (_estimator_)
3. Be clear about the assumptions that connect method to goal (_design_)
November 11, 2025 at 11:24 AM
I think the general lessons are:
1. Be clear about your goal (_estimand_)
2. Be clear about your analytical method (_estimator_)
3. Be clear about the assumptions that connect method to goal (_design_)
1. Be clear about your goal (_estimand_)
2. Be clear about your analytical method (_estimator_)
3. Be clear about the assumptions that connect method to goal (_design_)
3. This (fine, tolerable) paper by myself and Anna Grzymala-Busse.
What separates good from mere description? One key requirement is a correspondence between estimand and analysis -- good description is not bad causal inference.
github.com/ddekadt/good...
What separates good from mere description? One key requirement is a correspondence between estimand and analysis -- good description is not bad causal inference.
github.com/ddekadt/good...
github.com
November 11, 2025 at 11:24 AM
3. This (fine, tolerable) paper by myself and Anna Grzymala-Busse.
What separates good from mere description? One key requirement is a correspondence between estimand and analysis -- good description is not bad causal inference.
github.com/ddekadt/good...
What separates good from mere description? One key requirement is a correspondence between estimand and analysis -- good description is not bad causal inference.
github.com/ddekadt/good...
2. This fantastic paper by Brandon Stewart and Arthur Spirling
What are we doing when we run a regression, and how can we align that basic statistical tool with the thing we are meant to be doing -- explaining the world.
What are we doing when we run a regression, and how can we align that basic statistical tool with the thing we are meant to be doing -- explaining the world.
What Good Is a Regression? Inference to the Best Explanation and the Practice of Political Science Research | The Journal of Politics: Vol 87, No 4
We argue that almost all empirical social science research should employ a mode of argumentation called “Inference to the Best Explanation” (IBE). While elements of IBE appear widely, it is seldom con...
www.journals.uchicago.edu
November 11, 2025 at 11:24 AM
2. This fantastic paper by Brandon Stewart and Arthur Spirling
What are we doing when we run a regression, and how can we align that basic statistical tool with the thing we are meant to be doing -- explaining the world.
What are we doing when we run a regression, and how can we align that basic statistical tool with the thing we are meant to be doing -- explaining the world.
1. This brilliant but unfortunately-titled article by Miguel Hernán
The language you use to describe your research should be driven by the _estimand_ not the _estimator_: If you are studying a causal question, use causal language. But then expect people to evaluate the assumptions accordingly.
The language you use to describe your research should be driven by the _estimand_ not the _estimator_: If you are studying a causal question, use causal language. But then expect people to evaluate the assumptions accordingly.
The C-Word: Scientific Euphemisms Do Not Improve Causal Inference From Observational Data | AJPH | Vol. 108 Issue 5
Causal inference is a core task of science. However, authors and editors often refrain from explicitly acknowledging the causal goal of research projects; they refer to causal effect estimates as asso...
ajph.aphapublications.org
November 11, 2025 at 11:24 AM
1. This brilliant but unfortunately-titled article by Miguel Hernán
The language you use to describe your research should be driven by the _estimand_ not the _estimator_: If you are studying a causal question, use causal language. But then expect people to evaluate the assumptions accordingly.
The language you use to describe your research should be driven by the _estimand_ not the _estimator_: If you are studying a causal question, use causal language. But then expect people to evaluate the assumptions accordingly.
If I were a mysterious yet handsome young gentleman raised in a deeply class conscious society and desperately in want of a wife I would simply distinguish myself by having £10,000 a year.
November 11, 2025 at 8:27 AM
If I were a mysterious yet handsome young gentleman raised in a deeply class conscious society and desperately in want of a wife I would simply distinguish myself by having £10,000 a year.
All the money is in edgelord substacks these days
November 7, 2025 at 9:37 AM
All the money is in edgelord substacks these days
Just what I was hoping for!
November 6, 2025 at 8:30 AM
Just what I was hoping for!
So you’re saying Biden 2028!??
November 5, 2025 at 5:11 AM
So you’re saying Biden 2028!??