David Casacuberta points us to this story. Just throw in the Hoover Institution somewhere and we’d be all set!
9 thoughts on “Sleep, cats, and statistical significance: This one hits the Statistical Modeling, Causal Inference, and Social Science sweet spot!”
Actually, it’s better than most of the studies discussed here: “Nakahashi explained that the quantifiable data shows no significant difference.” An honest conclusion, with no p-hacks or forking paths in sight!
The linked post, from a blog focusing more on design than science, does a remarkably good job conveying the content of the study. The quantitative analysis is not fetishized and the qualitative results are reported as speculative. At the same time, I was engaged and informed while reading (though anything involving cats will have that effect on me).
The post loses nothing—and gains quite a lot—from abandoning the Standard Media Science Template. It might be an interesting exercise for the reader to come up with how the press release would look if this had been published in PNAS…
She slept where the cat chose to sleep, but selection of a place to sleep sounds like a forking path to me. For example, she and her cat can pick many places to sleep, but the cat might tend to pick some places harder for the research to fit due to size. What if she were to repeat the study, this time picking her own place to sleep? That might answer the question whether cats pick better sleeping places or variety in sleeping places is the key causal factor.
“selection of a place to sleep sounds like a forking path to me.”
You misread the article, and it’s WAAAAAAY worse than you think. She has FIVE cats. That’s a five-way fork every flipping night. That’s an octopus’s garden of forking paths.
Actually, it’s better than most of the studies discussed here: “Nakahashi explained that the quantifiable data shows no significant difference.” An honest conclusion, with no p-hacks or forking paths in sight!
Agreed!
The linked post, from a blog focusing more on design than science, does a remarkably good job conveying the content of the study. The quantitative analysis is not fetishized and the qualitative results are reported as speculative. At the same time, I was engaged and informed while reading (though anything involving cats will have that effect on me).
The post loses nothing—and gains quite a lot—from abandoning the Standard Media Science Template. It might be an interesting exercise for the reader to come up with how the press release would look if this had been published in PNAS…
No cat pic?
This one unfortunately has no statistics: N=1 https://www.cmaj.ca/content/169/9/940.short
But it did generate a rejoinder:
https://www.cmaj.ca/content/170/11/1646.2
She slept where the cat chose to sleep, but selection of a place to sleep sounds like a forking path to me. For example, she and her cat can pick many places to sleep, but the cat might tend to pick some places harder for the research to fit due to size. What if she were to repeat the study, this time picking her own place to sleep? That might answer the question whether cats pick better sleeping places or variety in sleeping places is the key causal factor.
Half :-)
Problems would remain.
First, the recruitment process is obviously problematic. All kinds of potential sources of bias.
Second, I suspect that there could be problems passing the internal review if someone else is telling the cat where to sleep?
“selection of a place to sleep sounds like a forking path to me.”
You misread the article, and it’s WAAAAAAY worse than you think. She has FIVE cats. That’s a five-way fork every flipping night. That’s an octopus’s garden of forking paths.
But she got the last laugh here: that paper got her into the best graduate school in Japan!
Here’s a video (in Japanese, but the cats speak cat, so it’s worth watching): https://tinyurl.com/58tx7b7m
XKCD is right on target today.
https://xkcd.com/2599/
Bob76