My argument has been that they're the best way to do hypothesis-driven science where there is risk of bias. This slide has been in every long-form RR talk I've given since 2013.pic.twitter.com/wpE7a1FW5x
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
I did? I don't recall saying that. From memory, you did preface a question about whether RRs could prevent oversampling & deliberate data selection (which led to an interesting pt about p-hacking) by saying that RRs wouldn't be right for your work, and I didn't question that
Modelling (the kind we are talking about here) doesn't play the role that inferential statistics play. You can't create a model by p-hacking, for example, since there are no p-values, right?
Looking at the data and then creating the model is fine — I mean maybe even required. Unless you explicitly label the model as having been tested on the data is was used to be trained, for example, I don't think it's a problem.
This is why I am sceptical about the value of pre-reg and RR, not because I don't think they don't work in the appropriate context, but because I think the debate needs to move forward into nuance.
Don't disagree with any of that. Just because prereg/RRs are a hammer doesn't make everything a nail. Let the nuanced debate flourish.
Yes, you agreed with me, I think. Didn't mean to imply you didn't.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.