This component contains all the files necessary for estimating the meta-analysis models reported in the paper.
The files are:
- `trippas_drh_syll.txt`: Trial-level data from all experiments. Coulmns are:
- `sub`: participant id. Unique within and across experiments
- `cond`: Experiment id. For mapping of experiment number to experiment see Table 2 in the paper.
- `conddesc`: description of experiment
- `syllc`: syllogistic form
- `val`: validity
- `bel`: believability
- `rsp`: binary response
- `rating`: six-point rating response
- `fit_meta-analysis_models.R`: `R` script that fits all meta-anylsis models. It requires the following files:
- `prep_data_trialwise.R`: Prepares the raw data for fitting.
- Model files as described below.
- `simulated.data.rda`: A simulated data set generated from fits of the original model for which the sigma-ratio is 1.5. This file is produced by `recover_generated_pars.R`
- `recover_generated_pars.R`: Generates one set of simulated data equivalent to the original data with the sole difference that the sigma-ratio is 1.5. Requires the following files:
- `fixed_models_Fms_1.Rda`: Results file from fitting the original model to the data. This file is not part of the current component.
- `prep_data_trialwise.R`: see above.
- `analysis_functions.R`: contains functions for analysis (e.g., plotting functions that are used here).
- model file of original model as described below (i.e., `models/uvsdt_6_fixed-criteria_sep-F_mu+sigma.stan`).
The `models` subfolder contains `.stan` files containing the model code. Note that parameter names in the model files do not necessarily correspond to the parameter names in the paper. What the parameters correspond to should be clear from either the model or the results scripts. In case things remain unclear contact Henrik Singmann (singman@gmail.com). The model files are (note that the files names are not particularly descriptive):
- `uvsdt_6_fixed-criteria_sep-F_mu+sigma.stan`: The 'original' meta-analyis model referred to throughout the paper.
- `uvsdt_6_fixed-criteria_sep-F_mu+sigma_alt.stan`: The meta-anylsis model with alternative prior specification.
- `uvsdt_6_fixed-criteria_weighted5_agg.stan`: The model without item effects (no $\eta^x_s$ model)
- `uvsdt_6_fixed-criteria_weighted5_agg-full.stan`: Fully aggregated model without participant- and item-effects (no $\xi^x$, no $\eta^x_s$ model). In contrast to the previous models this does not implement a random-effects meta-analysis.
Running the main script `fit_meta-analysis_models.R` will produce the results files (`.Rda` files) which are quite large. For example, the results file for the original model has a size of 3.2 GB. The fitted files are in the 'Fitted Meta-Analysis Component'.
