# Introduction --------------- This project includes all data used in the paper, as well as the source code used for behavioral analysis, and for modelling. To run the source code, the following software is required (installation should take not more than 30min): - R (preferably the latest version). - The following R packages: Rcpp, RcppEigen, RcppParallel, Rmalschains, rstan, dplyr, reshape2, entropy, glmmTMB, lsr, effectsize, ggplot2 For behavioral analysis run the **behavioral analysis.R** code Modeling requires several stages: A) use the stan code named **modelForCreatingAssociationSpace.stan** to create the association space. An example script demonstrating how to do this is provided in **ratingsModel.R**. This code can take ~6 hours to run, depending on the computer (tested on a standard windows PC with 8 virtual cores). An already estimated space of associations used in the study can be found in **forModelling.Rdata**. B) Use the estimated space of associations to fit the different models, by using the file SMPfitting.R. This file depends on the model which is coded using Rcpp (SMPmodel.cpp for Models 0, 1 and 3, and SMPmodel_twoRhos.cpp for Model 2). This code can take ~12 hours to run (for each model) for 80 participants, on a standard PC with 8 virtual cores. The code has been tested on windows PC, 32GB RAM, and 8 virtual cores, using R 4.0.2. # Files ------------- * data_forBehavioralAnalysis.csv - data file used for all raw RT analysis * behavioral analysis.R - code used for analyzing raw RT * modelForCreatingAssociationSpace.stan - the STAN code for the model used to create the association space * ratingsModel1.R - code for running the estimation of association spaces (depends on modelForCreatingAssociationSpace.stan) * optionalAdditionalDataForAssociationSpaces.csv - ratings given for trials used as attention checks (established strong and weak pairs of associates, where the second word was not an association given by the participant). These can be added to the Bayesian models to get a better estimation of individual-level ratings * forModelling.Rdata - a list, per subject, of all variables needed for the SMP modeling * SimulateFromSMP.R - code used to simulate responses+RTs from the SMP * SMPfitting.R - code used to fit the SMP (based on forModelling.Rdata as data, SMPmodel.cpp as model code, and SimulateFromSMP.R for simulations used for parameter recovery) * SMPmodel.cpp - Rcpp, parallelized version of the SMP model (for Models 0,1 and 3) * SMPmodel_twoRhos.cpp - Rcpp, parallelized version of the SMP model with separate rho for the initial association, and for subsequent associations (Model 2) * questionnaires.csv - questionnaire responses association spaces to improve the modeling of AS ratings # Variables in data (data_forBehavioraAnalysis.csv) ---------------- * Cue - cue for which an association should be given * Reaction.Time - RT in ms * Normed_associations - list of associations given to the cue by participants in Nelson et al., 2006 * Task_Name - Generation-Rep for the suppress group, and Generation-Rep-OK for the control group * pTP - distribution of the proportion of participants in Nelson et al., 2006 giving each association * pTP_entropy - the entropy of the pTP distribution * pTP_entropyGroup - whether entropy is high or low * Trial.Number.X - the index of the trial (doesn't start from 1 because training trials were excluded) * subject - subject index * Score - 100 for correct responses, 0 - for responses that did not appear in previous norms and therefore were counted as incorrect; -99 - for repeated associations (such scores should not appear in the control group) * Repeated1 - whether is a repeated association (given to the same cue) * rate - rating given by the participant to the association (0-100) * isWord - whether is an English word based on the qdapDictionaries library * isWord1 - whether is an English word based on SUBTLEX-UK or SUBTLEXus frequency lists * TPresponse - the pTP of the response the participant has given (NA if a non-normed association). The actual associations were redacted due to ethical considerations