Ingo Rohlfing
@ingorohlfing.bsky.social
I am here for all interesting and funny posts on the social sciences, broadly understood and including open science and meta science, academia, teaching and research. https://linktr.ee/ingorohlfing
Reposted by Ingo Rohlfing
"If we are unable to educate clinicians then merely persuading them to use CI's rather than p-values is to replace
the unthinking use of one technique with that of another."
A.P.Grieve (1992) Royal Statistical Society News and Notes, 18(7), 3-4.
the unthinking use of one technique with that of another."
A.P.Grieve (1992) Royal Statistical Society News and Notes, 18(7), 3-4.
November 18, 2025 at 6:29 PM
"If we are unable to educate clinicians then merely persuading them to use CI's rather than p-values is to replace
the unthinking use of one technique with that of another."
A.P.Grieve (1992) Royal Statistical Society News and Notes, 18(7), 3-4.
the unthinking use of one technique with that of another."
A.P.Grieve (1992) Royal Statistical Society News and Notes, 18(7), 3-4.
I read this post yesterday again because I was reading on many websites about CIs as covering the population value with 95% (or so) probability. I started to doubt my understanding of CIs and needed reassurance.
November 19, 2025 at 6:52 AM
I read this post yesterday again because I was reading on many websites about CIs as covering the population value with 95% (or so) probability. I started to doubt my understanding of CIs and needed reassurance.
Same here. (Incidentally, we cover confidence intervals tomorrow in Stats 101.) However, when one says "Only use confidence intervals" (not you, but others), you are mandating a certain research interest, which is what one may or may not have in a given study.
November 18, 2025 at 2:53 PM
Same here. (Incidentally, we cover confidence intervals tomorrow in Stats 101.) However, when one says "Only use confidence intervals" (not you, but others), you are mandating a certain research interest, which is what one may or may not have in a given study.
Reposted by Ingo Rohlfing
Luckily, you don't need to choose b/c you can use them both! The framing of "p value bad"/"CI good" means that some people never really get that they're based on the same information & the same basic stat. philosophy (some of these points are made here: richarddmorey.medium.com/power-and-pr...)
Power and precision
Why the push for replacing “power” with “precision” is misguided
richarddmorey.medium.com
November 18, 2025 at 1:42 PM
Luckily, you don't need to choose b/c you can use them both! The framing of "p value bad"/"CI good" means that some people never really get that they're based on the same information & the same basic stat. philosophy (some of these points are made here: richarddmorey.medium.com/power-and-pr...)
At least it's different, but valid point. p-values can be useful in some contexts and for some questions, confidence intervals for others. I never got why CIs should be superior to p-values per se. I would have to read ab New Statistics again, but I can't imagine a convincing reason.
November 18, 2025 at 10:23 AM
At least it's different, but valid point. p-values can be useful in some contexts and for some questions, confidence intervals for others. I never got why CIs should be superior to p-values per se. I would have to read ab New Statistics again, but I can't imagine a convincing reason.
I am sure they spent a lot of time thinking about what should and shouldn't be included in the interface. As they write, it reflects their experience with teaching quantitative methods, so I am sure it meets their needs.
November 14, 2025 at 2:44 PM
I am sure they spent a lot of time thinking about what should and shouldn't be included in the interface. As they write, it reflects their experience with teaching quantitative methods, so I am sure it meets their needs.
actual data, probably messy, to answer real-world questions.
The tabs seem a bit overloaded to me with input elements, data, formulas and plots, but this is just my personal impression.
Here is the direct link to the website: 2k1.iq.harvard.edu 2/
The tabs seem a bit overloaded to me with input elements, data, formulas and plots, but this is just my personal impression.
Here is the direct link to the website: 2k1.iq.harvard.edu 2/
November 14, 2025 at 12:57 PM
actual data, probably messy, to answer real-world questions.
The tabs seem a bit overloaded to me with input elements, data, formulas and plots, but this is just my personal impression.
Here is the direct link to the website: 2k1.iq.harvard.edu 2/
The tabs seem a bit overloaded to me with input elements, data, formulas and plots, but this is just my personal impression.
Here is the direct link to the website: 2k1.iq.harvard.edu 2/
I am not sure either whether this is for real, a scam or something else. The listed organizers are actual people, at least, and they refer to this agents4science.stanford.edu as their role model, which looks legit, but who can tell these days?
November 12, 2025 at 2:45 PM
I am not sure either whether this is for real, a scam or something else. The listed organizers are actual people, at least, and they refer to this agents4science.stanford.edu as their role model, which looks legit, but who can tell these days?
this, but this is necessarily still in its infancy. If one combines three new uses of AI in one format, how do you which one works better or worse and under what conditions? If an LLM tells you an LLM-written paper on LLM-based qualitative research is great, would you buy this w/o human scrutiny? 3/
November 12, 2025 at 1:53 PM
this, but this is necessarily still in its infancy. If one combines three new uses of AI in one format, how do you which one works better or worse and under what conditions? If an LLM tells you an LLM-written paper on LLM-based qualitative research is great, would you buy this w/o human scrutiny? 3/