Main content

Home

Menu

Loading wiki pages...

View
Wiki Version:
The performance data frame extracted from the ATC simulator logs is available in experiment_data.csv, and in img/dats_raw.RData. 1-extract.R contains the code that was used to extract this data from the (very large) raw XML logs. This script does not need to be run, as the extracted data is already provided in the repository. In the extracted data (experiment_data.csv and img/dats_raw.RData), the relevant columns are ppt (participant number), sess (session number), block (block number within a session), cond (experimental condition), stimulus (was the presented aircraft pair a conflict - c- or nonconflict - n), failtrial (did the automation make an incorrect recommendation, i.e., fail, on that trial, true or false), RT (response time in milliseconds), and R (did the participant respond conflict - C - or non-conflict - N). The 'response', 'score', and 'cumulative score' columns can be ignored, they are internal variables from the simulator that were only extracted for checking purposes. All the paper's analyses are included in scripts numbered 2-5. Note that before the data was analysed, non-responses and RTs <200ms were removed (as described in the manuscript). This 'cleaned' data is provided in img/dats_clean.RData. Scripts 2 & 2.5 (standard analyses) stand alone just run from this data. LBA analyses (scripts 3-5) were performed with a bespoke version of the dynamic models of choice R suite. The particular version used is provided in the 'dmc' submodule of this repo. Standard dmc releases can be found at https://osf.io/pbwx8/, along with lessons and more information about the software. Script 3-model_specification.R creates the LBA models reported in the paper. The most important model (reported in text) is the first one, which is stored in CA_top_samples. After models are created they must be sampled to obtain posterior distributions of parameter estimates. R/grid_dispatch.R dispatches sampling on a grid system running pbs pro. Sampling could be performed on an individual computer by using h.RUN.dmc (note that run time would likely be overnight or longer). Once the model is fitted, the modeller should create a subdirectory samples_data/, and the samples object should be stored inside. Other objects, such as posterior predictive data, will also be stored in samples_data in later scripts (4-5). A fitted copy of the posterior samples for the primary model of inference is provided (CA_top_samples.RData). Samples for the models discussed in the supplementary materials are also provided.