3/3 My personal highlight of #ESCOP2025 was becoming this year‘s #WomWom research fairy! 🧚‍♂️ Thanks again to all my fellow #WomeninWorkingMemory. It means a lot ❤️

Thanks for creating this! Would be great if you would add me 😇

We had some weird effects where the evidence for an interaction was ambiguous with dummy coded factors, but with the same data and sum-to-zero coded factors there was strong evidence for an interaction based on Bayes Factors.

Yes, the priors are typically set in the effect coefficients. Thus, dummy coding changes the prior predictive for different group means. As the reference group would only get the intercept prior, whereas an effect group would get intercept prior + effect prior.

For frequentist analysis the coding might be more of a convenience choice, depending on which coefficients are easier to interpret for you. But in Bayesian analyses the coding can make a big difference for the inference you draw. At least when using Bayesian Factors & prior predictive comparisons.

For details on the Memory Meausurement Model (M3) see this article: venpopov.github.io/bmm/articles...

And for all #decisionmaking researchers, you can easily use the model for any multinomial data assuming continuous activation strengths underlying the competitive selection among candidates

For anyone who ist familiar with R package structure I actually think it’s a very neat way to include a code book for your data and also adequately document any functions you might use.

If there’s is any questions with respect to the analysis feel free to reach out ☺️

Thank you so much for this positive feedback ❤️ And this is the first paper where I have used the R package structure to document and organize the code. I really enjoyed the features of roxygen documentation for datasets and analysis.