Peder M Isager

🚨 SynthNet is out 🚨
Researchers propose new constructs and measures faster than anyone can track. We (@anniria.bsky.social @ruben.the100.ci) built a search engine to check what already exists and help identify redundancies; indexing 74,000 scales from ~31,500 instruments in APA PsycTests. 🧵1/3

Regardless, this issue does pose problems for the usefulness of RV. I discuss this some in my PhD thesis discussion. See figure 6.3.C for a similar example: pure.tue.nl/ws/portalfil...

Test severity may at once improve the ability of a replication study to reduce uncertainty, but it can also leave less uncertainty to be reduced to begin with.

”paradox: test severity seems to imply both increased and reduced uncertainty reduction, depending on how you think about it”
I think this is not a paradox per se. Rather, I think test severity can have two causal effects that work against each other.

I think it is possible to reduce uncertainty about such claims through replication, even without a SESOI (e.g. if you’re willing to be a Bayesian).

”we need a SESOI and an equivalence-test framework”.
On the one hand I agree, but on the other hand, many scientific claims are not related to a SESOI. The claim is simply ”drug A works better than drug B”, ”there is a positive correlation between X and Y”, and so on.

Thanks for sharing your thoughts, this is definitely a worthwhile problem to consider. Let me share some unfinished thoughts of my own:

The post is a light and practical introduction to causal inference with variable control. It assumes som familiarity with regression analysis, and it helps to read my previous posts on causal inference, but that is it. Links to more in-depth literature on the topic are provided in references.

YouTube video by Peder Isager Which variables to control for, and why

I am currently experimenting with AI-augmentation of my educational materials. This post includes a NotebookLM video summary for those out there who prefer video tutorials to written summaries: youtu.be/EYTNzfHmTvc. The video does not replace the post, but provides a decent high-level summary.

and why adjusting for the wrong ones can introduce new, serious biases. The post features a worked example in Jamovi using simulated data. All files available on OSF >

Which variables to control for, and why | Peder M. Isager Personal website of Dr. Peder M. Isager

New blog post! ”Which variables to control for, and why”: pedermisager.org/blog/which-v...
In this post I give a beginner-friendly introduction to causal inference and statistical control, explaining why adjusting for the right variables clarifies relationships >

Join the next PMGS workshop where @ambra-prg.bsky.social will explain how to use Quarto for your reproducible research! 🔓🧑‍🔬

Thanks to all the commentary authors for your contributions, and thanks especially to the editors at Meta-psychology especially for allowing us to test out this format for our publication. It was a long road, but the end result is in my opinion terrific!

Quantifying replication value as a combination of citation count and sample size is our first stab at solving a very complex problem. Such early attempts benefit enormously from being critiqued right away. We were really lucky to receive several excellent commentaries from many experts in the field.

This paper forms one of Meta-Psychology’s Special Topics. Eight commentaries have been published alongside the paper which criticizes and extends the ideas presented within. We have written a response to these commentaries here: open.lnu.se/index.php/me...

My paper with @lakens.bsky.social and @annaveer.bsky.social - “Replication value as a function of citation impact and sample size” - has just been published in Meta psychology! open.lnu.se/index.php/me...

me with some garden hoses connected in a X -> Z <- Y fashion. If I shut the valve at Z, water from X spills out at Y

I built a DAG diagram with garden hoses for teaching.
Pictured: a collider bias diagram, inspired by a blocked pipe situation I experienced (which I credit with giving me the intuition though it also ruined my belongings in the flooded cellar).

Indeed, I'm looking forward!

Extremely honored to recieve Oslo New University College's science award for 2025. ONH has been a fantastic base to conduct my research at for the past 4 years, and I have an amazing team of colleagues around me to thank for that. From the bottom of my heart, thank you all!

Okay everyone, things are getting serious. I’m going to teach research methods again, 1st year psychology undergraduates. What would you cover with respects to philosophy of science, the research process etc.? I’m not very happy with the textbook stuff so I’m open to all ideas!

True, and maybe that should be emphasized in the blog post. Will consider adding a section at the end of the post. Any sources you'd recommend I'd cite in a section like this?

An abbreviation (ABB) in a journal article (JA) or Grant Application (GA) is rarely worth the words it saves. Every ABB requires cognitive resources (CR) and at my age by the time I'm halfway through a JA or GA I no longer have the CR to remember what your ABB stood for.

To be clear, the point of this post is not to say only experiments support causal inference and correlational research never can. Quite the opposite in fact. If that was your takeaway, I may need to add some language to clarify.

The motivation for picking the example in this post was simply that I wanted a thought experiment that a 1st year undergraduate in most areas of social and health science could wrap their head around without any additional reading.

I actually use the history of smoking~cancer research as my running example when introducing our bachelor students to research design and experimental vs observational research. It's a terrific example, albeit on a tragic subject.

Yes, one of several practical problems. You have to assume a perfect implementation for the hypothesis diagram to be true without caveats. Still, coming from a field where none of this is made explicit in most textbooks, I think understanding the unrealistically perfect case can be helpful.

Given that I wanted to cap the post at ~1000 words a lot of nuance will obviously be lost. Still, I think this post provides a useful preface to a proper causal inference introduction like Pearl's primer book.

Thanks for the reference, will check this out! I absolutely agree, and I don't think I say anything of the sort. However, I wanted to write a post that in very simple terms lays out the causal graphical logic underlying the textbook statement "you have to use experiments to make causal claims".

Later on the authors recommend abductive inference. I agree with the recommendation. However, abduction that leads us to a causal conclusion is just causal inference by another name. Saying that causal inference is not allowed is not sage advice. Better to emphasize that causal inference is hard.

I understand that sensible causal inference based on (cross-sectional) correlations can be hard. Very hard. But that does not mean our goal should not be causal inference. If a network modeling approach invalidates any causal inference, I question the usefulness of such a modeling approach.