Matt Williams
@matthewmatix.bsky.social
Associate prof at Massey University. Interested in statistics, open science, meta-psychology, and conspiracy theories. https://mattwilliams.netlify.app/
In which the authors of a 2023 Psych Science paper claim that the correlation between two constants is 1
October 10, 2025 at 1:34 AM
In which the authors of a 2023 Psych Science paper claim that the correlation between two constants is 1
What an odd paper!
This bit sounds faintly reasonable at first glance, but do we have any reason to expect that cross-sectional mediation would be much use for identifying candidate mediators to test experimentally? And how can something "account for an association" without having causal effects?
This bit sounds faintly reasonable at first glance, but do we have any reason to expect that cross-sectional mediation would be much use for identifying candidate mediators to test experimentally? And how can something "account for an association" without having causal effects?
August 21, 2025 at 9:35 AM
What an odd paper!
This bit sounds faintly reasonable at first glance, but do we have any reason to expect that cross-sectional mediation would be much use for identifying candidate mediators to test experimentally? And how can something "account for an association" without having causal effects?
This bit sounds faintly reasonable at first glance, but do we have any reason to expect that cross-sectional mediation would be much use for identifying candidate mediators to test experimentally? And how can something "account for an association" without having causal effects?
Pause to think: Do you have evidence of causality? Or does your discussion, perhaps, say that your study is correlational and cannot be used to draw causal inferences?
If so, have a hmm.
If so, have a hmm.
July 8, 2025 at 12:44 AM
Pause to think: Do you have evidence of causality? Or does your discussion, perhaps, say that your study is correlational and cannot be used to draw causal inferences?
If so, have a hmm.
If so, have a hmm.
This lines up with something we'd written in an earlier paper:
February 18, 2025 at 9:54 PM
This lines up with something we'd written in an earlier paper:
The TL;DR? On average, increased belief in one conspiracy theory *does* seem to lead to increased belief in others. But the effects were very small!
February 18, 2025 at 9:53 PM
The TL;DR? On average, increased belief in one conspiracy theory *does* seem to lead to increased belief in others. But the effects were very small!
Footage from the life of the mid-career researcher, 2024
October 30, 2024 at 11:32 PM
Footage from the life of the mid-career researcher, 2024
October 4, 2024 at 1:35 AM
I'm looking into an article on EMDR for depression with some numeric inconsistencies & implausible effect sizes. Wondered how it might be contaminating the literature so checked out a meta-analysis. Behold, all ye, this forest plot:
emdr-belgium.be/wp-content/u...
emdr-belgium.be/wp-content/u...
September 24, 2024 at 9:22 AM
I'm looking into an article on EMDR for depression with some numeric inconsistencies & implausible effect sizes. Wondered how it might be contaminating the literature so checked out a meta-analysis. Behold, all ye, this forest plot:
emdr-belgium.be/wp-content/u...
emdr-belgium.be/wp-content/u...
If there's one* thing I hate about ResearchGate, it's how they plaster prominent "Request file pdf" buttons on articles that are *open access* by just clicking the DOI link (even though their system picks up when a paper has CC-BY licence!). So I regularly get file requests for OA papers.
February 16, 2024 at 2:16 AM
If there's one* thing I hate about ResearchGate, it's how they plaster prominent "Request file pdf" buttons on articles that are *open access* by just clicking the DOI link (even though their system picks up when a paper has CC-BY licence!). So I regularly get file requests for OA papers.
It's reassuring to know that history, economics and politics has nothing to do with gender inequality. Take that, historians
December 20, 2023 at 9:45 AM
It's reassuring to know that history, economics and politics has nothing to do with gender inequality. Take that, historians
Discussion: The relationships are mediated by sex hormones. Did we even measure those? No, of course not.
December 20, 2023 at 9:31 AM
Discussion: The relationships are mediated by sex hormones. Did we even measure those? No, of course not.
HOW DID YOU KNOW
December 20, 2023 at 9:21 AM
HOW DID YOU KNOW
Yes but they added years of schooling to the national IQ estimates so it's all fixed
December 20, 2023 at 9:01 AM
Yes but they added years of schooling to the national IQ estimates so it's all fixed
This is mainly a note to myself, but in case anyone else cares: lavaan gives two different estimates of the robust RMSEA and CFI.
The better set to use are the ones with their own separate rows (cfi.robust and rmsea.robust), also called the sample corrected versions.
The better set to use are the ones with their own separate rows (cfi.robust and rmsea.robust), also called the sample corrected versions.
October 25, 2023 at 12:30 AM
This is mainly a note to myself, but in case anyone else cares: lavaan gives two different estimates of the robust RMSEA and CFI.
The better set to use are the ones with their own separate rows (cfi.robust and rmsea.robust), also called the sample corrected versions.
The better set to use are the ones with their own separate rows (cfi.robust and rmsea.robust), also called the sample corrected versions.
Attaining sufficient power for interaction analyses appears to be far beyond the resource constraints of the typical study. In many circumstances we might be better off just not running interaction analyses at all...
October 3, 2023 at 8:23 PM
Attaining sufficient power for interaction analyses appears to be far beyond the resource constraints of the typical study. In many circumstances we might be better off just not running interaction analyses at all...