Thanks so much for the answer! Well, I wrote that we would stop if the observed effect was under the SESOI, but this rule did not impact the alpha spending calculation.

Thanks!

At peek 1, my interaction is smaller than the SESOI. But decomposing it shows the expected pattern, only smaller than anticipated. I decide to collect extra data despite being below SESOI. At peek 2, interaction is significant at the adjusted alpha level.

Does this increase the type I error rate?

link.springer.com/article/10.1...

We're used to international comparisons involving less than 50 data points. Here's a county level analysis with > 2k data points.

For each data point, the author calculated a proportion of tweets supporting conspiracy theories, and examined its relation with inequalities indicators. It's mental.🤯

I'm very grateful to the chief editors Professors Shelley McKeown Jones (‪@shelleymckeown.bsky.social‬) and Sammyh Khan for asking me to fulfil the position!

*allowed to joke about bribery, not allowed to take bribes. 2/2

You can listen to the complete soundtrack on Youtube: youtu.be/f73OPXgj4_k?...

#indiegame #OST #coffee #gaming

🧐👉 This paper is available in #OA: doi.org/10.5334/irsp...

#SocialPsyc
#OpenScience

@paulbertin.bsky.social
@kenzonera.bsky.social

Thank you for sharing, Mark!

For instance, I owe much of my modest understanding of Bayesian probabilities to a combination of readings + ChatGPT which really facilitated my learning.

But even for the things ChatGPT rocks at, it requires to master the basics enough to be able to notice when it makes mistakes.

Cool, let's organize a coffee expedition then :D I'll write to you tomorrow.

You're welcome! If you're available sometime in the upcoming weeks, I'd love to chat about a paper I'm currently work on (it involves light bayesian belief update modeling, and I'd like to have the thoughts of someone more competent than me on the issue). I'll write you an email.

SH*T THAT IS YOUUU!