# README The files in this repository accompany the paper: Brochhagen (2021): *Brief at the risk of being misunderstood: Consolidating population- and individual-level tendencies*. Computational Brain & Behavior. DOI: 10.1007/s42113-021-00099-x https://dx.doi.org/10.1007/s42113-021-00099-x ---- ### Commented directory * `full-amb-run.R` reads data from `data/` and feeds it to the $5$ models in `models/`. Results are saved in `fits/` * `get-loo-rmse.R` reads model fits from `fits/`, runs diagnostics to rule out pathologies related to HMC (`stan_utility.R`) and then outputs models' leave-one-out cross-validations, their root mean squared error, and model comparisons for models fit on the same data. * `models/` contains the Stan models (see Appendix A for details) * `fp1-amb-model.stan` is $\text{FullPool}_{\lambda}$ * `fp2-amb-model.stan` is $\text{FullPool}_{\lambda, \alpha,\beta}$ * `np-amb-model.stan` is $\text{NoPool}$ * `pp-amb-model.stan` is $\text{HM}_{\lambda}$ * `ppf-amb-model.stan` is $\text{HM}_{\lambda,\alpha,\beta}$ * `stan_utility.R` defines functions to diagnose pathologies. This is a minimally modified version of [Michael Betancourt's](https://betanalpha.github.io/writing/) script by the same name. See also: http://mc-stan.org/users/documentation/case-studies/rstan_workflow.html ---- Note that, in order to run the `full-amb-run.R`, the data from [Kanwal et al (2017)](http://dx.doi.org/10.1016/j.cognition.2017.05.001) needs to put into the appropriate place & format. In more detail: The script expects the data to be located in `data/`. It requires: (i) a $\text{subject} \times \text{object_to_convey}$-matrix assigned to `obs_mtx`; (ii) a $\text{subject} \times \text{message_sent}$-matrix assigned to `prod_mtx`; (iii) a $\text{subject} \times \text{outcome_freq_obj}$-matrix, that tracks success (+1) or failure (-1) communicating the frequent object using the short form to convey the frequent meaning, assigned to `cnts0_mtx`; and (iv) a $\text{subject} \times \text{outcome_infreq_object}$, analogous to (iii) but for the infrequent object, assigned to `cnts1_mtx`. For instance, if using all the data, all matrices should have $40$ rows (subjects) and $32$ columns (trials). The raw data is available at: http://datashare.is.ed.ac.uk/handle/10283/2702