You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Tired of covid news? Here's some replication crisis commentary. :) My lab read "Science is not a signal detection problem" by Wilson, Harris, & Wixted: https://www.pnas.org/content/117/11/5559 … An interesting defense of the role of NHST in scientific practice, but I was unconvinced. [thread]pic.twitter.com/BFfMZEWkqU
WHW argue that the "failure" of OSC 2015 (1/3 successful replications) is due to regression to the mean, under the assumption that the literature is filtered for p < .05.
Isn't that saying that the replication crisis is "just due to publication bias"? Yeah but... 
The article goes on to defend a publication bias ecosystem in which discovery science cheaply "filters" effects to find candidate large magnitude papers. But this scenario is exactly what is critiqued in Button et al.'s new classic on "power failure":https://www.nature.com/articles/nrn3475 …
From the perspective of decision theory, the "live with pub bias" view seems badly flawed. In my lab + ManyLabs + ManyBabies, it takes 10-100x as much time and effort to debug or debunk a published p < .05 finding that is incorrect or artifactual. e.g.: https://doi.org/10.1016/j.jesp.2015.10.012 …
So setting up incentives that encourage publication of low-quality "candidate effects" seems like a recipe for disaster. Based on the cost of multi-site replication, most findings will not be replicated, and hence will live in the (increasingly junky) literature as zombies.
Further, the general scientific model proposed by the article moves from "signal detection" (effect present or zero) to "effect finding" (continuous effects). But - to paraphrase Newell (1973) - you can't play twenty effect sizes with nature and win. https://mindhacks.com/2015/02/10/you-cant-play-20-questions-with-nature-and-win/ …pic.twitter.com/iBgJEv6c7N
The article also suggests that large sample sizes are irresponsible. This seems totally wrong. The more carefully you measure an effect, the better able you are to detect variation that informs theory. Our work on child language is an example of this: https://langcog.github.io/wordbank-book/conclusion-beyond-cdi.html#methodological-morals …pic.twitter.com/yVGsNdVSFh
Overall, this piece reifies a number of negative aspects of the current scientific ecosystem and suggests that they might actually be adaptive. I much prefer reform proposals that center around the role of precise measurement and theory building.
Thanks for this much needed commentary
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
